OFAI

FuC LLMs 2026 -- The Future of Computational Linguistics in the Age of LLMs

2026-03-04T00:00:00+01:00

In March 2026, OFAI will host a select group of researchers discussing the future of computational linguistics in the age of AI. The discussions will include but will not be restricted to questions such as: Is computational linguistics (CL) dead, or is there still a place for CL beside LLMs? Do Transformer-based LLMs or multi-modal models bring us closer to real natural language understanding (NLU), or do we need a different architecture for true progress? Do LLMs reduce everything to a linguistic problem?

The symposium is co-organised by Brigitte Krenn (OFAI) and Stephanie Evert (Friedrich-Alexander-Universität Erlangen-Nürnberg).

Location:

Österreichisches Forschungsinstitut für Artificial Intelligence (OFAI) Freyung 6/6, 1010 Wien, Austria, 3rd floor (no elevator!)

Schedule:

Wed 04.03. afternoon: positioning & goals (starting with position statements from participants)
Thu 05.03.: brainstorming & discussions (based on key questions)
Fri 06.03.: summary, position statement, action points
Sam 07.03.: informal debriefing, review of the symposium results, and next steps (who: the organizers and all participants still present and interested)

Gleichbehandlungs-Blog Beitrag zu KI-Ethik und Diskriminierungsbekämpfung

2026-01-27T00:00:00+01:00

OFAI's Brigitte Krenn hat mit der Gleichbehandlungsanwaltschaft im Rahmen deren Gleichbehandlungs-Blogs über die ethische Nutzung von KI sowie die Bekämpfung von Diskriminierung mithilfe von KI gesprochen und welche Rolle die Daten dabei spielen. Hier geht es zum "Ethik und Diskriminierungsbekämpfungs"-Beitrag.

Gleichbehandlungs-Blog Beitrag zu Simulation oder echte künstliche Intelligenz

2026-01-26T00:00:00+01:00

OFAI's Brigitte Krenn hat mit der Gleichbehandlungsanwaltschaft im Rahmen deren Gleichbehandlungs-Blogs über was ist KI eigentlich, was meinen wir, wenn wir von AI oder KI sprechen, und welche kollektiven Fehlkonzeptionen haben wir bezüglich was AI kann, was es ist, oder eben nicht kann und nicht ist. Hier geht es zum "Simulation oder echte KI"-Beitrag.

JOKER workshop at CLEF 2025

2025-10-21T00:00:00+02:00

Papers from the workshop "JOKER: Automatic Humour Analysis" have now been published in the proceedings of CLEF 2025, the 16th Conference and Labs of the Evaluation Forum.

The JOKER workshop, co-organized by OFAI's Tristan Miller, aims to bring translators and computer scientists together to develop data and evaluation metrics for the (semi-)automatic translation of humorous language. The workshop consists of a number of shared tasks, held in the spring of 2025, followed by a series of sessions at the CLEF 2025 conference in Madrid from 9 to 12 September 2025.

A peer-reviewed overview of the entire workshop appears in Experimental IR Meets Multilinguality, Multimodality, and Interaction, a volume in Springer's Lecture Notes in Computer Science series. Papers describing the individual shared tasks of the workshop, as well as the participating systems, are published in Working Notes of CLEF 2025, a volume in the CEUR Workshop Proceedings series.

Full bibliographic details of the OFAI-coauthored papers are as follows:

Liana Ermakova, Ricardo Campos, Anne-Gwenn Bosser, and Tristan Miller. Overview of the CLEF 2025 JOKER lab: Humour in the machine. In Jorge Carrillo de Albornoz, Julio Gonzalo, Laura Plaza, Alba García Seco de Herrera, Josiane Mothe, Florina Piroi, Paolo Rosso, Damiano Spina, Guglielmo Faggioli, and Nicola Ferro, editors, Experimental IR Meets Multilinguality, Multimodality, and Interaction: 16th International Conference of the CLEF Association, CLEF 2025, Madrid, Spain, September 9–12, 2025, Proceedings, volume 16089 of Lecture Notes in Computer Science (ISSN 0302-9743), pages 315–337, Cham, Switzerland, September 2025. Springer. ISBN 978-3-032-04353-5. DOI: 10.1007/978-3-032-04354-2_18.
Liana Ermakova, Ricardo Campos, Anne-Gwenn Bosser, and Tristan Miller. Overview of the CLEF 2025 JOKER task 1: Humour-aware information retrieval. In Guglielmo Faggioli, Nicola Ferro, Paolo Rosso, and Damiano Spina, editors, CLEF 2025 Working Notes, CEUR Workshop Proceedings, pages 2744–2760. CEUR-WS.org, 2025.
Liana Ermakova, Anne-Gwenn Bosser, Tristan Miller, and Ricardo Campos. Overview of the CLEF 2025 JOKER task 2: Wordplay translation from English into French. In Guglielmo Faggioli, Nicola Ferro, Paolo Rosso, and Damiano Spina, editors, CLEF 2025 Working Notes, CEUR Workshop Proceedings, pages 2761–2779. CEUR-WS.org, 2025.
Liana Ermakova, Tristan Miller, Yaël Naud, Anne-Gwenn Bosser, and Ricardo Campos.
Overview of the CLEF 2025 JOKER task 3: Onomastic wordplay translation. In Guglielmo Faggioli, Nicola Ferro, Paolo Rosso, and Damiano Spina, editors, CLEF 2025 Working Notes, CEUR Workshop Proceedings, pages 2780–2790. CEUR-WS.org, 2025.

New Project on the Scientific Preparation of a Federal Large-Scale Language Model (Bundes-LLM)

2025-10-01T00:00:00+02:00

Brigitte Krenn is starting a new R&D service project: Scientific preparation of a federal large-scale language model.

The study addresses technological and societal aspects (legal, ethical and ecological) of various cases of the implementation of a Bundes-LLM, including: foundation model training and adaptation, fine-tuning of existing LLMs (third party models or one's own base model) to the application needs of individual stakeholders, LLM-based platforms such as RAG systems, and their adaptation to the application needs of individual stakeholders. Sovereignty-First Implementations entirely based on-premises infrastructure, with all data processed and stored within national borders and Hybrid Flexibility Models where sensitive data are processed on-premises, while public services are delivered through a secure cloud environment, and Ecosystem Partnerships based on the collaboration with European AI initiatives to share resources and expertise, reducing R&D costs and accelerating capability transfer.

The outcome of the study is designed to provide a robust, lawful and ethical foundation for decision-makers regarding the development and deployment of a Bundes-LLM for public administration, also taking into account the potential for applications in industry and businesses. The study will clarify key questions and inform strategic choices regarding professional requirements engineering for diverse user groups, ensuring data and technology sovereignty through legal, sustainable and ethical integration, project development, management, and operation in the public interest, cost efficiency through opensource components. Roadmaps for the different usecases and policy implementation recommendations are provided.

For more details see the project website (in German).

Launch of the GermanDialects-AI project

2025-09-02T00:00:00+02:00

We are happy to announce the official launch of the project German Dialects: Document, Preserve and Learn with the help of AI.

The project officially kicked off with a meeting on September 2nd. This collabotative effort brings together OFAI, LMU Munich and the University of Liechtenstein, serving as lead partner.

The German Dialects AI project aims to document and preserve Upper German dialects, promoting their learning via an AI-enhanced online platform.

We look forward to the exciting developments and insights this partnership will bring!

Interspeech Paper

2025-08-19T00:00:00+02:00

INTERSPEECH 2025 took place in Rotterdam where OFAI was represented by Lorenz Gutscher and Michael Pucher. Their paper "Audio-Based Classification and Geographic Regression of Austrian Dialects" was presented at the conference and is now available in the ISCA archive.

Read the paper in the ISCA archive

Article: Austrian Dialect Classifier

2025-08-01T00:00:00+02:00

Lorenz Gutscher and Michael Pucher had an interview with Jana Wiese - a journalist for futurezone.at and Kurier. The outcome is a nice article that is easy to understand and describes what we have worked recently: An AI model that predicts the location from your dialect speech. It was printed in the newspaper Kurier and published online in an extended version at futurezone.at Link: Article: Futurezone.at

Try it out yourself and record your speech to see where the AI locates you: Link to Huggingface

Article: Austrian Dialect Classifier

2025-08-01T00:00:00+02:00

Try it out yourself and record your speech to see where the AI locates you: Link to Huggingface

Citizen Science Award

2025-04-16T00:00:00+02:00

We’re excited to announce that our project DialektDetect has been nominated for the Citizen Science Award by the OeAD!

After training an AI model to predict where in Austria a dialect recording originates, we’re turning the tables — can humans do it better? In this interactive project, school classes and youth groups are invited to listen to short audio clips of dialect speakers and guess their origin on a map. Participants will also highlight the part of the recording that influenced their decision the most.

Take part online and compete for prize money of 1,000€, 750€, or 500€! To participate with your class/youth group write an email to lorenz.gutscher@ofai.at to receive your class code, which you will need to take part in the challenge.

Link: demo.ofai.at/DialektDetect

Further information: youngscience.oead.at – DialektDetect

KI und Stimmen @ Bundes-Blindeninstitut

2025-04-11T00:00:00+02:00

While Text-to-Speech systems might seem like a convenient gadget to some, they can be life-changing for blind or visually impaired individuals.

In a recent workshop at the Bundes-Blindeninstitut Wien, we presented our ongoing work in speech synthesis. Participants also joined our Citizen Science Award project DialektDetect, where they listened to dialect recordings and guessed the speaker’s origin on a map — relying solely on audio cues.

The highlight was to test wheather the AI can correctly predict their own dialect by recording themselves.

DialektDetect @ Wiener Forschungsfest 2025

2025-03-23T00:00:00+01:00

In this citizen science project, we explore how well humans and AI can identify Austrian dialects based solely on audio. Participants placed location markers on an Austrian map, competing against the prediction of the AI.

It was impressive to see how some children instantly recognized dialects that sounded like those of their classmates or relatives from specific regions. Not only could they pinpoint the origin, but they also identified key characteristics of the dialects. For unknown regions, localization was a challenge for old and young.

The project successfully provided citizens with an accessible and practical way to engage with AI, while also demonstrating that AI is not infallible and can make mistakes.

OFAI's presentation of LEGO Audio & Braille Building Instructions at the Wiener Forschungsfest 2025 - a big success!

2025-03-23T00:00:00+01:00

OFAI presented LEGO Audio & Braille Building Instructions at the Wiener Forschungsfest 2025. In two and a half days approx. 400 visitors built small LEGO models blindfolded, only with the help of the audio instructions. Since the Forschungsfest was targeted to young people offering a wide range of experiences in the world of science, our stand was besieged by children, their parents, and interested adult visitors.

Our project LEGO Audio & Braille Building Instructions was initially targeted to blind and visually impaired LEGO fans, providing them with verbal instructions such they can build their models independently. This not only demands a much higher cognitive load, since one has to conceptualize first which brick is to be searched for and then how to precisely mount it on a given structure without seeing it. But through the lack of eye-hand coordination, significantly higher fine-motoric skills are needed as well. Some comments from visitors corroborate our ideas that the range of applications could actually much broader.

We could observe visitors (mostly kids) that did fantastically well, and others that had a real challenge, but all of them reported that it is indeed much more difficult. They moved on with an experience of the world of the blind, but also with a feeling of success that they were able to build a model without seeing it. We were also quite challenged (on Sunday, there were 6 of us, constantly setting up new builds and monitoring the progress), and we left the Forschungsfest tired, happy, and with rich experiences and insights.

OFAI is represented at the Wiener Forschungsfest 2025 with 2 contributions

2025-03-21T00:00:00+01:00

OFAI will provide 2 contributions to the Wiener Forschungsfest 2025 that takes place from March 21st -- 23rd in the localities of the Viennese townhall.

The project LEGO Audio & Braille Building Instructions is all about providing blind and visually impaired LEGO fans with verbal instructions such they can build their models independently either using audio or reading Braille text.

The Citicen Science Award OeAD project DialektDetect This project compares how spoken Austrian dialects are recognized by humans and how the process of recognition differs from that of a machine-trained model.

Can AI create humour?

2025-01-24T00:00:00+01:00

OFAI Associate Researcher Tristan Miller is interviewed for matrix - computer & neue medien, ORF Radio Ö1's weekly show on the digitalization of society. In the episode, "Kann Künstliche Intelligenz Humor?", he discusses PunCAT, a computer-assisted humour translation tool jointly developed with the University of Vienna's Centre for Translation Studies.

Peter Hallman speaks at the Workshop Cause(e/r) "The Interplay between Event Structure and Argument Realization"

2025-01-23T00:00:00+01:00

Peter Hallman spoke by invitation at the Workshop Cause(e/r): The Interplay between Event Structure and Argument Realization at the University of Cologne, January 23-24, 2025.

Organized by Haydar Batuhan Yildiz and Jens Hopperdietzel at the University of Cologne, the workshop explored the syntactic and semantic status of intermediary participants in complex causative constructions, such as causees or instruments, from a cross-linguistic perspective. Topics included argument hierarchies, (in)direct causation, agentivity, among others, as they manifest themselves in German, English, French, Arabic, Turkish and other languages.

OFAI's LEGO audio instructions at Munich school for the blind and visually impaired

2025-01-22T00:00:00+01:00

Successful test session of OFAI’s LEGO audio instructions at Munich school for the blind and visually impaired

LEGO accessibility platform manager Frederik Hansen and OFAI linguist Susanne Höfler met up in Munich to visit the Sehbehinderten- und Blindenzentrum Südbayern, an association housing schools and various day-care facilities for children and young adults with visual impairments.

The aim of the day: Putting their audio instructions for LEGO sets to the test. OFAI has been in charge of the LEGO Audio & Braille Building Instructions project since 2018, having released English instructions for over 100 models so far. That day, however, the team would test the waters with a new prototype: the first instructions in German.

A lively and eager group of seven students between the ages of 10 and 14 with varying degrees of visual impairment happily volunteered for a casual LEGO play session. After a short round of introductions, the kids split into three groups and were handed one tablet with instructions and one of three different LEGO models per group – they could pick between a green race car, a hot-dog truck and a cat playground. Most of our testers were quite experienced with building LEGO sets, but none of them had ever done so on the basis of LEGO audio instructions. Our team was eager to find out not only if the kids could smoothly build with the audio instructions but also if they would actually enjoy this new form of accessibility. The unanimous verdict after the play session was: Oh yes! And much to our delight, both fully blind children and children with considerably lower degrees of visual impairment were able to benefit from the audio instructions and engage in productive and enjoyable co-play using the different visual and auditory means provided in the instructions. The children not only helped us with providing a proof-of-concept, but they were virtually overflowing with suggestions and ideas on how to make the building experience for their target group more enjoyable. We’re very grateful for the children’s trust, time and enthusiasm and want to thank the Sehbehinderten- und Blindenzentrum Südbayern (PR and fundraising representative Stephanie Märkl in particular) as well as the children’s parents, caretakers and teachers for making this afternoon possible.

OFAI’s LEGO Audio & Braille Building Instructions project is all about providing blind and visually impaired LEGO fans with verbal instructions that can be accessed for free at legoaudioinstructions.com. The instructions come in three flavors: Users can access LEGO audio instructions directly via the website in a simple web UI (and if users need the help of a seeing person, all instructions are accompanied by the well-known LEGO images), download text instructions for Braille reader or access text instructions to be used with the screen reader of their choice.

Festveranstaltung 55 Jahre ÖSGK, 40 Jahre OFAI

2024-12-05T00:00:00+01:00

Jubiläumsveranstaltung des Österreichischen Forschungsinstituts für Artificial Intelligence (OFAI)

55 Jahre Österreichische Studiengesellschaft für Kybernetik und 40 Jahre Österreichisches Forschungsinstitut für Artificial Intelligence

Vor 40 Jahren haben sich österreichische WissenschafterInnen zusammengetan, um in einem universitätsunabhängigen Institut, dem "Österreichischen Forschungsinstitut für Artificial Intelligence (OFAI)" auf dem faszinierenden Gebiet der Artificial Intelligence zu forschen. Dies in zweifacher Hinsicht: Um Computer "intelligent" zu machen, damit sie unser Leben erleichtern, und um durch Computermodelle mehr über uns Menschen zu erfahren.

Das muss gefeiert werden! Die Jubiläumsveranstaltung findet statt im Festsaal in der Beletage des Palais Teuffenbach, am Donnerstag, den 5. Dezember 2024 um 16:00 Uhr.

Für Kurzentschlossene sind noch Restplätze vorhanden.

Peter Hallman receives the Mohammad bin Rashid Arabic Award

2024-11-25T00:00:00+01:00

OFAI Researcher Peter Hallman has been awarded the Mohammad bin Rashid Arabic Language Award in the category of ‘distinguished individual’ for the year 2024 at an awards ceremony in Dubai on October 10, 2024. The Mohammed Bin Rashid Arabic Language Award is the highest appreciation of the efforts of individuals and organizations for the Arabic Language, and is part of the initiatives launched by His Highness Sheikh Mohammed Bin Rashid Al Maktoum, Vice President and Prime Minister of the United Arab Emirates and Ruler of Dubai, to celebrate the pivotal role of the Arabic language in enriching global cultural dialogues, in addition to enhancing the status of the Arabic language and encouraging those who aim to nourish it. The category ‘distinguished individual’ honors an individual who has been instrumental in bringing Arabic into the purview of art, science, or technology internationally and illuminating Arabic on the world stage. Hallman received the award for his longstanding and intensive linguistic research on the Arabic language, especially in the underrepresented area of semantics. The awards ceremony was held in the context of the 10th Annual International Arabic Language Conference, where over 2000 participants from 85 countries attended talks on Arabic education, technology, media, communication, language policy, history, linguistics, and many other topics.

How AI is changing university education

2024-10-22T00:00:00+02:00

OFAI Associate Researcher Tristan Miller is interviewed for Dimensionen, ORF Radio Ö1's daily show on science and research. In the episode, "Wie KI die Uni-Lehre verändert", he speaks about the history, limitations, and educational role of AI, as well as his recent project on computer-assisted translation of humour.

JOKER workshop at CLEF 2024

2024-09-19T00:00:00+02:00

Papers from the workshop "JOKER: Automatic Humour Analysis" have now been published in the proceedings of CLEF 2024, the 15th Conference and Labs of the Evaluation Forum.

The JOKER workshop, co-organized by OFAI's Tristan Miller, aims to bring translators and computer scientists together to develop data and evaluation metrics for the (semi-)automatic translation of humorous language. The workshop consists of a number of shared tasks, held in the spring of 2024, followed by a series of sessions at the CLEF 2024 conference in Grenoble from 9 to 12 September, 2024.

A peer-reviewed overview of the entire workshop appears in Experimental IR Meets Multilinguality, Multimodality, and Interaction, a volume in Springer's Lecture Notes in Computer Science series. Papers describing the individual shared tasks of the workshop, as well as the participating systems, are published in Working Notes of CLEF 2024, a volume in the CEUR Workshop Proceedings series.

Full bibliographic details of the OFAI-coauthored papers are as follows:

Liana Ermakova, Anne-Gwenn Bosser, Tristan Miller, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. Overview of the CLEF 2024 JOKER track: Automatic humour analysis. In Lorraine Goeuriot, Philippe Mulhem, Georges Quénot, Didier Schwab, Laure Soulier, Giorgio Maria Di Nunzio, Petra Galuščáková, Alba García Seco de Herrera, Guglielmo Faggioli, and Nicola Ferro, editors, Experimental IR Meets Multilinguality, Multimodality, and Interaction: Proceedings of the Fifteenth International Conference of the CLEF Association (CLEF 2024), volume 14959 of Lecture Notes in Computer Science (ISSN 0302-9743), pages 165–182, Cham, September 2024. Springer. ISBN 978-3-031-71907-3. DOI: 10.1007/978-3-031-71908-0_8.
Liana Ermakova, Anne-Gwenn Bosser, Tristan Miller, and Adam Jatowt. Overview of the CLEF 2024 JOKER task 1: Humour-aware information retrieval. In Guglielmo Faggioli, Nicola Ferro, Petra Galuščáková, and Alba García Seco de Herrera, editors, Working Notes of the Conference and Labs of the Evaluation Forum (CLEF 2024), volume 3740 of CEUR Workshop Proceedings, pages 1775–1785, 2024.
Victor Manuel Palma Preciado, Grigori Sidorov, Liana Ermakova, Anne-Gwenn Bosser, and Adam Jatowt. Overview of the CLEF 2024 JOKER task 2: Humour classification according to genre and technique. In Guglielmo Faggioli, Nicola Ferro, Petra Galuščáková, and Alba García Seco de Herrera, editors, Working Notes of the Conference and Labs of the Evaluation Forum (CLEF 2024), volume 3740 of CEUR Workshop Proceedings, pages 1786–1799, 2024.
Liana Ermakova, Anne-Gwenn Bosser, Tristan Miller, and Adam Jatowt. Overview of the CLEF 2024 JOKER task 3: Translate puns from English to French. In Guglielmo Faggioli, Nicola Ferro, Petra Galuščáková, and Alba García Seco de Herrera, editors, Working Notes of the Conference and Labs of the Evaluation Forum (CLEF 2024), volume 3740 of CEUR Workshop Proceedings, pages 1800–1810, 2024.

10 to 13 September 2024 KONVENS Conference in Vienna

2024-09-10T00:00:00+02:00

OFAI as local co-organiser is pleased to announce the KONVENS (Konferenz zur Verarbeitung natürlicher Sprache/Conference on Natural Language Processing) which will take place in Vienna, Austria from September 10 to 13, 2024.

KONVENS is an annual conference series on computational linguistics that started in 1992 and is organized under the auspices of the German Society for Computational Linguistics and Language Technology, the Special Interest Group on Computational Linguistics of the German Linguistic Society, the Austrian Society for Artificial Intelligence and SwissText.

GerMS-Detect Workshop on Sexism Detection in German Online News Fora, at KONVENS2024 in Vienna

2024-09-10T00:00:00+02:00

We are pleased to announce the GerMS-Detect Workshop on Sexism Detection in German Online News Fora which will take place as part of the KONVENS conference, in Vienna, Austria on 10 September, 2024.

In the workshop, the results from the GermEval 2024 Shared Task 1 GerMS-Detect are presented and further discussed. The papers related to the workshop presentations will be made available from 10 September at the workshop page.

Podiumsdiskussion zu AI in der Medienproduktion

2024-09-10T00:00:00+02:00

Wir laden Sie herzlich zur Podiumsdiskussion "AI in der Medienproduktion" am 10. September 2024 um 16:30 im Bundesministerium für Klimaschutz, Umwelt, Energie, Mobilität, Innovation und Technologie (BMK) ein!

Die Teilnahme ist frei! Bitte bringen Sie einen Ausweis mit, um in den Veranstaltungssaal im BMK eingelassen zu werden!

Im Rahmen des GerMS-Detect Workshop on Sexism Detection in German Online News Fora bei der KONVENS 2024 findet am 10 September eine prominent besetzte Podiumsdiskussion zum Thema "AI in der Medienproduktion" statt. Es diskutieren Vertreter*innen von Medien, AI-Forschung und -Anwendung, sowie IT- und Medienrecht über den Einsatz von AI in der Medienproduktion, aktuelle Rahmenbedingungen und Auswirkungen, sowie Bedarfseinschätzungen und Visionen für die Zukunft. Details siehe hier (Deutsch) und hier (Englisch).

GerMS-AT Corpus Published

2024-07-31T00:00:00+02:00

The GerMS-AT (German Misogyny/Sexsim - Austria) Corpus, which has been used for the GermEval2024 GerMS-Detect shared task has now been fully released publicly under a very permissive license (CC-BY-NC-SA-4.0).

GerMS-AT Dataset

2024-07-31T00:00:00+02:00

The GerMS-AT (German Misogyny/Sexism - Austria) dataset contains user comments from an Austrian online newspaper. The comments have been annotated by 4 or more out of 11 annotators as to how strong sexism/mysogyny is present in the comment.

For each comment, the code of the annotator and the label assigned is given for all annotators which have annotated that comment. Labels represent the severity of any sexism/misogyny present in the comment from 0 (none), 1 (mild), 2 (present), 3 (strong) to 4 (severe).

The dataset contains 7984 comments. We provide the data using the same split as was used for the GermEval2024 GerMS-Detect shared task with a training set of 5998 comments and a test set of 1986 comments. No dev set is provided as the choice of dev set may be best left to the machine learning researcher/engineer.

A unique propery of this corpus is that it contains only a small portion of sexist/misogynyst remarks which use strong language, curse-words or otherwise blatantly offending terms, a large number of comments contain more subtle, indirect or at times ambiguous forms of sexism/misogyny.

Publications

Brigitte Krenn, Johann Petrak, Marina Kubina, and Christian Burger. 2024. Germs-at: A sex-ism/misogyny dataset of forum comments from an Austrian online newspaper. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 7728–7739.

New website for OSGK

2024-07-12T00:00:00+02:00

Our parent organization, the Austrian Society for Cybernetic Studies (OSGK), has a new website.

OFAI was established in 1984 as a special research institution of OSGK. OSGK, founded in 1969, is a non-profit organization of scientists and practitioners aiming to study the theoretical bases of cybernetics and related fields, including artificial intelligence.

Examining the sources of visual humour

2024-07-09T00:00:00+02:00

A computer-assisted study on the role of scale distortions in visual humour has been published in the European Journal of Humour Studies.

The article, "On the Use of Scale Distortion for Visual Humour: A Preliminary Analysis", examines human ratings of humour in a data set of cartoons where the size of various objects has been distorted. The analysis reveals that scenes with distorted objects are perceived to be significantly funnier than the original images.

The article is co-authored by Clara Swaboda, formerly of the University of Vienna, and OFAI associate researcher Tristan Miller.

Ibn Jinni and the origin of language

2024-07-08T00:00:00+02:00

In "Social consensus and divine inspiration", Peter Hallman looks at remarks on the origin of language by Muslim Arab linguist and philosopher Ibn Jinni. The article appears in the Summer 2024 issue of Babel: The Language Magazine.

em. O. Univ. Prof. Dr. Hubert Haider, Univ. Salzburg

2024-06-19T00:00:00+02:00

OFAI is proud to present "Multiple targets – Ambiguitäten als Kriterium für LLMs", a talk by Hubert Haider of University of Salzburg. The talk is part of OFAI's 2024 Spring Lecture Series (will be held in German).

Members of the public are cordially invited to attend the talk online via Zoom on Wednesday, 19 June 2024 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

Talk abstract: Laut AI-Index-Report 2024 wurde die menschliche Leistung als Baseline im Fall von Natural Language Inference bereits von LLMs überschritten. Es gibt aber einen sprachlichen Aufgabentyp, bei dem diese Systeme schlechter abschneiden. Das ist das Erkennen und Berücksichtigen syntaktischer Ambiguitäten, was der Vortrag zu zeigen sich bemühen wird.

Speaker biography: Hubert Haider ist emeritierter Prof. für Linguistik mit der Spezialisierung auf Grammatiktheorie (Syntax germanischer, romanischer und slawischer Sprachen; kognitive Evolution von Grammatiksystemen). Als Prof. an der Univ. Stuttgart engagierte er sich in den Anfängen der "maschinellen Sprachverarbeitung" im SFB 340 (Linguistic Foundations for Computational Linguistics) und als Leiter eines DFG-Graduiertenkollegs (Linguistic foundations for language processing). Mit dem Wechsel an die Univ. Salzburg wurde er Mitgründer und Mitglied des dortigen Centre for Cognitive Neuroscience.

Stephanie Gross - invited lunch lecture at the Department of Innovation and Digitalisation in Law

2024-06-13T00:00:00+02:00

Stephanie Gross gave an invited talk on "Bias in Language Models" at the Department of Innovation and Digitalisation in Law at the University of Vienna on June 13, 2024.

Dr. Margherita Pallottino, Univ. of Geneva / OFAI

2024-06-05T00:00:00+02:00

OFAI is proud to present "Fii uses across Arabic varieties", a talk by Margherita Pallottino of University of Geneva / OFAI. The talk is part of OFAI's 2024 Spring Lecture Series.

Members of the public are cordially invited to attend the talk online via Zoom on Wednesday, 05 June 2024 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

Talk abstract: The particle fii (etymologically the preposition ‘in’, with allomorph bi-) is universally attested across Arabic varieties. Alternative uses of this element have been identified in single dialects. For instance in the Levantine dialects fii is used in existential constructions (Jarad, 2015) or as a modal (Cowell, 1964), whereas in the North African region fii is an ingredient of progressive constructions (Brahim, 2007). The existing literature gives us a picture of the multifunctionality of fii in Arabic but it does not provide any indication about the geographical distribution of these usages, nor about the possible correlations among them. The research described in this talk fills this gap surveying 11 different uses of fii previously described in the literature, across 11 different Arabic varieties extending from Morocco to Iraq. The resulting picture shows a leopard-spot distribution of certain functions of fii suggesting that other forces drive the emergence of fii usages across varieties beside the influence of geographically proximity. The main question tackled by this talk is, therefore, which factors can explain the distribution of fii functions. The validity of the competing hypotheses, historical accident or grammar driven, are evaluated with mathematical methods run on the data gathered in the study.

Bibliography: Brahim, A. (2007). Le marquage locatif de l’objet et aspect progressif en arabe et en berbère tunisien. In Morphosyntaxe et sémantique du verbe. Relations actiancielles, voix, aspect et statut grammatical en français et en arabe (p. 94‑105). CRISCO. Cowell, M. W. (1964). A reference grammar of Syrian Arabic (Based on the dialect of Damascus). Georgetown University Press. Jarad, N. I. (2015). From locative to existential: The grammaticalization of fii in the spoken Arabic of Aleppo. Romano-Arabica, XV, 235‑254.

Speaker biography: Margherita Pallottino holds a PhD in Theoretical Linguistics from the University of Geneva. Her 2020 thesis, written under the direction of Prof. Tabea Ihsane and Prof. Luigi Rizzi, investigates the case marking function of the preposition fii in Tunisian Arabic. Margherita's research focuses on interdialectal variation among Arabic Varieties. She is particularly interested in the distribution of locative prepositions and of their use in a crosslinguistic perspective. In July 2022, Margherita joined the OFAI (Österreichisches Forschungsinstitut für Artificial Intelligence - Austrian Research Institute for Artificial Intelligence) as post-doc fellow. Her fellowship is sponsored by the FNS Mobility Grant P500PH_211169/1. At the OFAI Margherita is working at the creation of a browsable linguistic atlas describing the polyfunctionality of the prepositions bi and fii across twelve Arabic Dialects, spanning from Moroccan Arabic in the west to Iraqi Arabic in the east. The aim of the research project is the identification of patterns of grammaticalization which allow single prepositional elements to perform a wealth of syntactic functions in closely related languages.

Prof. James E. Young, BSc, PhD, University of Manitoba

2024-05-29T00:00:00+02:00

OFAI is proud to present "Designing Human–Robot Interaction", a talk by Prof. James E. Young, BSc, PhD of the University of Manitoba. The talk is part of OFAI's 2024 Spring Lecture Series.

Talk abstract: How should we interact with a robot? How can we give it commands? Get information from it? Robots' real world, often collocated and autonomous presence, provides a range of new and exciting opportunities for re-envisioning interaction with technology. In this talk, Dr. Young will present his team's work on exploring novel interaction with robots through a range of projects over the last 12 years. A key focus of this work is aiming to solve HRI problems through novel interaction design rather than technological advances, re-conceptualizing problems to make them simpler. Further, Dr. Young's team explores the limits of robots' abilities to use emotion and human social interaction techniques, for example, to deceive and manipulate people. Finally. Dr. Young will introduce his lab's current projects on re-designing domestic companion robot interactions with a focus on simplicity and deployability.

Speaker biography: Jim Young is a Professor of Computer Science at the University of Manitoba, where he founded the Human-Robot Interaction lab in 2011, and co-directs the Human-Computer Interaction Lab. Jim received his PhD in Computer Science from the University of Calgary in 2010, following a BSc in Computer Science from Vancouver Island University in 2005. Jim's work broadly takes a human-centric focus and draws heavily from sociology and psychology, focusing on studying social interactions between people and robots and inventing new ways for people to work with them. He currently serves as Editor-in-Chief for the ACM Transactions on Human-Robot Interaction Journal, Steering Committee co-chair for the ACM/IEEE International Conference on Human-Robot Interaction (HRI), and is on the steering committee for the ACM International Conference on Human-Agent Interaction (ACM HAI). He previously served as Program co-chair for the ACM/IEEE HRI 2017 and General co-chair for ACM/IEEE HRI 2020, as well as in many editorial roles including Senior Associate Editor for the ACM Transactions on Human-Robot Interaction journal, PC member and sub-theme chair for the ACM/IEEE HRI conference, and Associate Editor for the Frontiers In Robotics and AI (Human-Robot Interaction) journal. The HRI Lab's work has been recognized by best paper awards at various ACM Conferences.

Paper at LREC-Coling 2024

2024-05-23T00:00:00+02:00

OFAI and derStandard have coauthored a paper "GERMS-AT: A Sexism/Misogyny Dataset of Forum Comments from an Austrian Online Newspaper" whihc was presented at THE 2024 JOINT INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS, LANGUAGE RESOURCES AND EVALUATION (LREC-COLING 2024), 20-25 MAY, 2024 / TORINO, ITALIA.

The paper can be downloaded from ACL anthology.

Jordan Kodner, PhD, Stony Brook University

2024-04-24T00:00:00+02:00

OFAI is proud to present "Is it Language or Task Design? Reinterpreting language models' recent successes in morphology and syntax learning", a talk by Jordan Kodner of Stony Brook University, NY. The talk is part of OFAI's 2024 Spring Lecture Series.

Members of the public are cordially invited to attend the talk online via Zoom on Wednesday, 24 April 2024 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: The success of neural language models (LMs) on a wide range of language-related tasks may be in part due to their ability to induce human-like representations or understanding of natural language grammars. Humans are, after all, gold-standard language learners. For the past several years, researchers pursuing this question have developed a number of methodologies for testing the grammar representations learned by LMs that have reached generally positive conclusions. I will take a critical look at such studies in this talk. While modern LMs are clearly extremely impressive, and clearly do often capture important aspects of natural language grammars, the methodologies of many popular studies have unfairly overestimated the capacities of LMs when it comes to their ability to induce human-like representations. Focusing on questions of hierarchical syntactic representations and generalization in inflectional morphology, I will discuss how unintended biases in data-splitting, artificial training or test data, overly simplistic evaluations, weak or absent baselines, and faulty interpretations, have conspired to overestimate the abilities of LMs. While the conclusions of this study are largely negative in terms of the current state-of-affairs, they are also optimistic. By employing more thorough and rigorous methodologies, we have developed a better scientific understanding of the nature of LMs and representations of the grammar. In identifying weak points for current models, we points towards research areas where greater improvements may be gained.

Speaker biography: Jordan Kodner is an Assistant Professor in the Stony Brook University Department of Linguistics and an affiliate of the Institute for Advanced Computational Science and Natural Language Processing group. His primary research revolves around computational approaches to child language acquisition and their broader implications. In particular, algorithmic models of grammar acquisition, especially morphology, how those processes drive language variation and change, what insights they provide for low-resource NLP, and what they tell us about the intersection of (low-resource) NLP and cognitive science. In 2020, he received his PhD from the University of Pennsylvania Department of Linguistics, where he worked with Charles Yang and Mitch Marcus. Prior to that, he received a master's degree from the University of Pennsylvania Department of Computer and Information Science in 2018. From 2013 through 2015, he was an Associate Scientist in the Speech, Language, and Multimedia group at Raytheon BBN Technologies where he worked on defense and medical-related projects.

CfP: GermEval2024 GerMS-Detect - Sexism Detection in German Online News Fora

2024-04-19T00:00:00+02:00

We are pleased to announce the GermEval2024 GerMS-Detect shared task about Sexism Detection in German Online News Fora, to be published at KONVENS2024.

For details see the Competition Website

Robert Trappl zu Gast in der Ö1 Sendereihe IM GESPRÄCH

2024-04-19T00:00:00+02:00

Andreas Obrecht im Gespräch mit Robert Trappl über den KI-Hype seit ChatGPT; über KI-Ethik und die Tendenz in Europa in KI-Ethik zu investieren statt in KI-Entwicklung, wie z.B. in den USA oder in China; über die Möglichkeiten von KI Menschen mit Zusatzfähigkeiten auszustatten, was wir vermehrt in der aktuellen Kriegsführung sehen, aber glücklicherweise auch darin, dass Menschen KI-gestützt Bewegungsmöglichkeiten und Sinneserfahrungen (wie z.B. Sehen und Hören) wieder bekommen; und über viele Themen mehr. Nachzuhören ist das Gespräch unter IM GESPRÄCH KI-Pionier Robert Trappl.

OFAI 2024 Spring Lecture Series

2024-04-12T00:00:00+02:00

OFAI is delighted to announce its 2024 Spring Lecture Series, featuring an eclectic lineup of internal and external speakers.

The talks are intended to familiarize attendees with the latest research developments in AI and related fields, and to forge new connections with those working in other areas.

Lectures will take place at 18:30 Vienna time, usually every other Wednesday. All lectures will be held online via Zoom; in-person attendance at OFAI is also possible for certain lectures. Attendance is open to the public and free of charge. No registration is required.

Subscribe to our newsletter or our RSS feed, or bookmark this web page, to receive further details for the individual talks.

24 April 2024 at 18:30 CEST (UTC+2)

Jordan Kodner (Stony Brook University)

Is it Language or Task Design? Reinterpreting language models' recent successes in morphology and syntax learning

The success of neural language models (LMs) on a wide range of language-related tasks may be in part due to their ability to induce human-like representations or understanding of natural language grammars. Humans are, after all, gold-standard language learners. For the past several years, researchers pursuing this question have developed a number of methodologies for testing the grammar representations learned by LMs that have reached generally positive conclusions. I will take a critical look at such studies in this talk. While modern LMs are clearly extremely impressive, and clearly do often capture important aspects of natural language grammars, the methodologies of many popular studies have unfairly overestimated the capacities of LMs when it comes to their ability to induce human-like representations. Focusing on questions of hierarchical syntactic representations and generalization in inflectional morphology, I will discuss how unintended biases in data-splitting, artificial training or test data, overly simplistic evaluations, weak or absent baselines, and faulty interpretations, have conspired to overestimate the abilities of LMs. While the conclusions of this study are largely negative in terms of the current state-of-affairs, they are also optimistic. By employing more thorough and rigorous methodologies, we have developed a better scientific understanding of the nature of LMs and representations of the grammar. In identifying weak points for current models, we points towards research areas where greater improvements may be gained.

29 May 2024 at 18:30 CEST (UTC+2)

Jim Young, BSc, PhD (University of Manitoba)

Designing Human–Robot Interaction

How should we interact with a robot? How can we give it commands? Get information from it? Robots' real world, often collocated and autonomous presence, provides a range of new and exciting opportunities for re-envisioning interaction with technology. In this talk, Dr. Young will present his team's work on exploring novel interaction with robots through a range of projects over the last 12 years. A key focus of this work is aiming to solve HRI problems through novel interaction design rather than technological advances, re-conceptualizing problems to make them simpler. Further, Dr. Young's team explores the limits of robots' abilities to use emotion and human social interaction techniques, for example, to deceive and manipulate people. Finally. Dr. Young will introduce his lab's current projects on re-designing domestic companion robot interactions with a focus on simplicity and deployability.

How to attend: Attend online via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.

You can add this event to your calendar.

05 June 2024 at 18:30 CEST (UTC+2)

Dr. Margherita Pallottino (University of Geneva / OFAI)

Fii uses across Arabic varieties

The particle fii (etymologically the preposition ‘in’, with allomorph bi-) is universally attested across Arabic varieties. Alternative uses of this element have been identified in single dialects. For instance in the Levantine dialects fii is used in existential constructions (Jarad, 2015) or as a modal (Cowell, 1964), whereas in the North African region fii is an ingredient of progressive constructions (Brahim, 2007). The existing literature gives us a picture of the multifunctionality of fii in Arabic but it does not provide any indication about the geographical distribution of these usages, nor about the possible correlations among them. The research described in this talk fills this gap surveying 11 different uses of fii previously described in the literature, across 11 different Arabic varieties extending from Morocco to Iraq. The resulting picture shows a leopard-spot distribution of certain functions of fii suggesting that other forces drive the emergence of fii usages across varieties beside the influence of geographically proximity. The main question tackled by this talk is, therefore, which factors can explain the distribution of fii functions. The validity of the competing hypotheses, historical accident or grammar driven, are evaluated with mathematical methods run on the data gathered in the study.

19 June 2024 at 18:30 CEST (UTC+2)

em. O. Univ.-Prof. Dr. Hubert Haider (Universität Salzburg)

Multiple targets – Ambiguitäten als Kriterium für LLMs

Laut AI-Index-Report 2024 wurde die menschliche Leistung als Baseline im Fall von Natural Language Inference bereits von LLMs überschritten. Es gibt aber einen sprachlichen Aufgabentyp, bei dem diese Systeme schlechter abschneiden. Das ist das Erkennen und Berücksichtigen syntaktischer Ambiguitäten, was der Vortrag zu zeigen sich bemühen wird.

How to attend: Attend online via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.

Article in GLOSSA Journal of General Linguistics

2024-04-05T00:00:00+02:00

View Peter Hallman's latest article, on parallels between causative and double object constructions and what they say about the base thematic hierarchy. Be sure to check out the supplementary file comparing the the notion of VP-shells to neoconstructionism.

Paper at Nature Scientific Reports

2024-03-28T00:00:00+01:00

Marcin Skowron has coauthored the article "Song lyrics have become simpler and more repetitive over the last five decades" published in Nature Scientific Reports (open access).

JOKER workshop previewed at ECIR 2024

2024-03-20T00:00:00+01:00

JOKER-2024, an upcoming workshop and shared task on Automatic Wordplay Analysis, will be previewed next week at the 46th European Conference on Information Retrieval (ECIR 2024) in Glasgow, Scotland.

The JOKER workshop series, co-organized by OFAI Associate Researcher Tristan Miller, aims to bring translators and computer scientists together to develop data and evaluation metrics for the (semi-)automatic translation of humorous language. JOKER-2023 will be held during the CLEF 2024 conference in Grenoble from 9 to 12 September, 2024.

The presentation of JOKER-2024 at ECIR 2024 is accompanied by the following paper in the conference proceedings:

Liana Ermakova, Anne-Gwenn Bosser, Tristan Miller, Tremaine Thomas-Young, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. CLEF 2024 JOKER lab: Automatic humour analysis. In Nazli Goharian, Nicola Tonellotto, Yulan He, Aldo Lipani, Graham McDonald, Craig Macdonald, and Iadh Ounis, editors, Advances in Information Retrieval: 46th European Conference on Information Retrieval, ECIR 2024, Glasgow, UK, March 24–28, Proceedings, Part VI, volume 14613 of Lecture Notes in Computer Science (ISSN 0302-9743), pages 36–43, Cham, March 2024. Springer. ISBN 978-3-031-56072-9. DOI: 10.1007/978-3-031-56072-9_5.

Mag. Dr. Brigitte Krenn and Mag. Dr. Stephanie Gross, MSc, OFAI

2024-03-13T00:00:00+01:00

Please note, originally this talk was scheduled for 14 February 2024 as part of OFAI's 2023 Fall Lecture Series. For this talk had to be rescheduled until Wednesday, 13 March 2024.

OFAI is proud to present "Bias in Language Models Illustrated by the Example of Gender", a talk by Brigitte Krenn and Stephanie Gross of the Austrian Research Institute for Artificial Intelligence.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 13 March 2024 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Bias in language models is a widely discussed topic in AI. So far, many voices have been raised calling for baises to be prevented in training data. Which, however, is a futile endeavor in many real-world contexts. In the present talk, the speakers argue for a different strategy which can be summarized from an AI ethics point of view as "be aware of and transparent about your desired and undesired biases". The feasibility and technical viability of such an approach will be illustrated in the talk. In particular, experiments in fine-tuning pretrained language models with gender-biased data are presented and the resulting outcomes are qualitatively and quantitatively analysed.

Speaker biography: Brigitte Krenn is Deputy Director of the Austrian Research Institute for Artificial Intelligence (OFAI). She has worked in natural language processing and AI since 1990. Her overall research interest lies in understanding and computationally modelling human language capability. She is board member of the Austrian Society for Artificial Intelligence (ASAI) where she heads the Working Group on Natural Language Processing.

Stephanie Gross is a research scientist at the Austrian Research Institute for Artificial Intelligence (OFAI). She has been involved as PI and Co-PI in different national and international research projects, focusing on the development, implementation, and analysis of AI-based technical systems, including quantitative as well as qualitative approaches. Her main research interests lie in the fields of natural language processing, including large language models, task-based multi-modal human-human and human-robot interaction, and language learning.

Peter Hallman invited talk at Uni Graz

2024-03-06T00:00:00+01:00

"Aspectual Composition in the Arabic Imperfective"

Peter Hallman gave an invited talk on "Aspectual Composition in the Arabic Imperfective" in the Workshop "Exploring Verbal Paradigms: Confronting Arabic and Slavic Languages in Tense and Aspect Realization", hosted at the Institute for Slavic Studies, University of Graz, on March 6, 2024.

Talk Abstract: (The slides of the talk can be accessed here.)

The Standard Arabic imperfective verb form has three “moods”, called “indicative”, “subjunctive”, and “jussive”. The indicative conveys the two meanings typically associated with imperfectivity cross-linguistically, namely the progressive and habitual aspects, but the other two moods are harder to pin down semantically. In this talk, I present evidence that the subjunctive corresponds to the English infinitive and has no meaning of its own. Rather, in the subjunctive form, the lexical aspect of the underlying verb shines through. I claim that the indicative mood also has no meaning of its own, but only applies to verb phrases that are aspectually stative. In order for an eventive verb to be put in the indicative, then, it must first be stativized by a null habitual or progressive operator, explaining the morphological uniformity of habitual and progressive aspects in Arabic and potentially cross-linguistically. I will have less to say about the jussive, but suggest that it bears a closer resemblance to the subjunctive than to the indicative and may be an allomorph of it.

Peter Hallman invited talk at University of Nova Gorica

2024-02-29T00:00:00+01:00

Peter Hallman gave an invited talk on "Reduced Clausal Comparatives in Arabic and Slavic" at the Center for Cognitive Science of Language, University of Nova Gorica, Feb. 29, 2024.

For more information see the talk abstract and the slides for the talk.

New FWF Top Citizen Science Project on "Dialect classification by human and artificial intelligence"

2024-02-06T00:00:00+01:00

The project titled “Dialect classification by human and artificial intelligence” led by Michael Pucher and Lorenz Gutscher has been approved as one of the 6 FWF Top Citizen Science Projects.

The primary objective is to develop an Explainable Artificial Intelligence (XAI) system dedicated to dialect classification. The system will identify crucial frequency regions and features, offering transparency in its decision-making process. As part of the project, participants are encouraged to engage with a user-friendly web interface, allowing them to record and analyze their own dialect. Moreover, the citizens’ knowledge will be used to test if human classification draws attention to similar regions as an XAI system does. This gives citizens a point of contact with the current (often overwhelming) development of AI systems and connects them with scientific research.

Michael Pucher/Lorenz Gutscher: Invited Talk at University of Zürich

2024-02-01T00:00:00+01:00

OFAI's Michael Pucher and Lorenz Gutscher gave a talk titled "Acoustic language embeddings and phonetic typology of Austrian German varieties" at the workshop "Vowel and consonant quantity in Germanic, Indo-European and beyond" at the University of Zürich.

Assoz. Prof. Dipl.-Ing. Dr. techn. Clemens Heitzinger, TU Wien

2024-01-31T00:00:00+01:00

Reinforcement learning has been instrumental in many advances in AI, including medicine. In such applications, statements about the reliability of the results are necessary in addition to convergence results. Research in this direction is the topic of "Reinforcement Learning and its Application in Medicine and Large Language Models", a talk by Clemens Heitzinger of TU Wien. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 31 January 2024 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Reinforcement learning has been instrumental in many advances in AI in recent years. The most publicized is certainly the development of ChatGPT and large language models (LLM) in general; the last and crucial training step of ChatGPT is reinforcement learning with human feedback (RLHF). Still, in order to fully solve learning problems, statements about the reliability of the results are necessary in addition to convergence results. For example, reliability and trustworthiness of AI systems is of utmost importance in medicine and other safety critical areas. In this talk, reinforcement-learning algorithms for training LLM and for calculating optimal treatments of sepsis patients are described. The questions of convergence to an optimal policy and of reliability are addressed by PAC (probably approximately correct) estimates and other approaches to policy evaluation.

Speaker biography: Clemens Heitzinger is Co-Director of the Center for Artificial Intelligence and Machine Learning (CAIML) at TU Wien and Associate Professor in the Department of Computer Science (Informatics) at TU Wien. He received both his master's degree (Dipl.-Ing.) in applied mathematics and his PhD degree (Dr. techn.) in technical sciences with highest honors from TU Wien. He was a visiting researcher in the Department of Mathematics and Statistics at Arizona State University, a research associate in the School of Electrical and Computer Engineering at Purdue University, and a senior research associate in the Department of Applied Mathematics and Theoretical Physics (DAMTP) at Cambridge University. In 2015, he returned to TU Wien as an associate professor. He is also Adjunct Professor in the School of Mathematical and Statistical Sciences at Arizona State University. He was awarded the START Prize, Austria's most prestigious award for young scientists, by the Austrian Science Fund (FWF) in 2013. He is author of Algorithms with Julia (Springer, 2022). His research interests are reinforcement learning and uncertainty quantification (in particular Bayesian inversion) with applications in the sciences, medicine, and engineering.

Univ.-Prof.in Dipl.-Psych. Dr.in Stefanie Höhl, University of Vienna

2024-01-17T00:00:00+01:00

In adults, the brain rhythms of interaction partners entrain to communicative rhythms, including speech, supporting mutual comprehension and communication. Whether this is also the case in the infant brain is the subject of "Social Rhythms and Biobehavioral Synchrony in Early Human Development", a talk by Stefanie Höhl of the University of Vienna. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 17 January 2024 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Caregiver-infant interactions are characterized by interpersonal rhythms at different timescales, from nursery rhymes and interactive games to daily routines. These rhythms make the social environment more predictable for young children and facilitate interpersonal biobehavioral synchrony with their caregivers. In adults, the brain rhythms of interaction partners entrain to communicative rhythms, including speech, supporting mutual comprehension and communication. I will present recent evidence that this is also the case in the infant brain, especially when babies are addressed directly by their caregiver through infant-directed speech in naturalistic interactions. Through using simultaneous measures of neural and physiological rhythms, e.g., dual-fNIRS and dual-ECG, from caregiver and infant during live face-to-face interactions, we can further deepen our understanding of early interactional dynamics and their reciprocal nature. I will present our recent research identifying factors supporting the establishment of caregiver-infant neural synchrony, such as affectionate touch and vocal turn-taking. I will further discuss the functional links and dissociations between caregiver-infant synchrony on the neural and physiological levels. I will outline potential implications of this work and point out important future directions.

Speaker biography: Stefanie Höhl is head of the Research Unit of Developmental Psychology at the University of Vienna where she leads the Wiener Kinderstudien lab. She received her PhD from the University of Leipzig and completed her Habilitation at the University of Heidelberg. From 2016 to 2019 she led the Max Planck Research Group on Early Social Cognition at the MPI for Human Cognitive and Brain Sciences in Leipzig. Her research at the intersection of developmental psychology and cognitive neuroscience focuses on social and cognitive development in early childhood. Grounded on an interactionist perspective, she applies EEG and fNIRS hyperscanning in caregiver-child and caregiver-infant interactions to study the neural dynamics of early social exchanges.

Happy Birthday Robert Trappl

2024-01-16T00:00:00+01:00

On January 16 we celebrated the 85th anniversary of Robert Trappl the founder and director of OFAI. See also the APA SCIENCE press release (in German) on the occasion of his birthday.

New Book "Introduction to Digital Humanism" now available in print and online

2024-01-12T00:00:00+01:00

The new “Introduction to Digital Humanism” with a contribution by Brigitte Krenn is now also available in print. The online version available since 22 December 2023 is already a big success with its approx. 80.000 downloads by the date of this posting.

New Book on Digital International Relations

2024-01-10T00:00:00+01:00

Brigitte Krenn at Diplomatische Akademie Wien, acting as facilitator for the book presentation „Digital International Relations: Technology, Agency and Order“ edited by Corneliu Bjola and Markus Kornprobst.

Prof. Mag. phil. Thomas Graf, MA, PhD, Stony Brook University

2023-12-20T00:00:00+01:00

Given the rapid rise of large language models, will symbolic linguistics be left in the dust, or is this actually an opportunity for meaningful synergy between symbolic and subsymbolic approaches? This question is addressed in "Linguistics and Symbolic Computation in a World of Large Language Models", a talk by Thomas Graf of Stony Brook University. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk via Zoom on Wednesday, 20 December 2023 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Language has always played a central role in artificial intelligence, yet AI researchers and linguists have rarely seen eye to eye on things, in particular the status of subsymbolic/neural approaches to language. After decades of debates, it looks like the subsymbolic approaches have finally emerged victorious. Not only are large language models (LLMs) succeeding in incredibly complex real-world tasks, subsymbolic models are also rapidly gaining traction in some areas of theoretical linguistics, e.g. lexical semantics. This raises the question: will symbolic linguistics be left in the dust, or is this actually an opportunity for meaningful synergy between symbolic and subsymbolic approaches?

In this talk, I argue for the latter by presenting “subregular syntax” as a concrete example of what such a synergy may look like. Subregular syntax is a symbolic approach that combines formal language theory with the Minimalist syntax framework proposed by Noam Chomsky, which grants it a large degree of empirical coverage across a wide range of typologically diverse languages. Despite that broad coverage, subregular syntax is a very simple formalism that analyzes all syntactic dependencies in terms of relativized adjacency conditions. Even though these conditions are stated over trees, they can actually be reduced to a very specific types of n-grams over strings. This opens up a new way of representing sentence structure in neural networks while bringing robust learning algorithms like stochastic gradient descent to Minimalist syntax. It also casts doubt on claims in the literature that the behavior of neural networks in specific linguistic tasks, e.g. binding or NPI-licensing, shows that they use tree structure. Instead, these findings may be indicative of a network’s ability to use fairly elaborate types of n-grams. The careful study of the symbolic approach of subregular syntax thus is an opportunity to deepen our understanding of neural networks while also harnessing their advantages for theoretical linguistics.

Speaker biography: Thomas Graf is Associate Professor of Computational Linguistics in the Department of Linguistics at Stony Brook University. He also holds an affiliate appointment in Stony Brook’s Institute for Advanced Computational Science. Before joining Stony Brook, he studied linguistics at the University of Vienna and received his PhD from UCLA in 2013. His research operates at the intersection of theoretical linguistics, computational linguistics, and cognitive science, with a particular focus on syntax (sentence structure). He is the recipient of the 2014 E.W. Beth Dissertation Prize for outstanding PhD theses in logic, language, and information, and in 2019 he received an NSF CAREER award for his research on subregular syntax. He is the creator of the blog Outdex, which covers computational and theoretical linguistics, and he loves to introduce high school students to the wonders of computational linguistics during Stony Brook’s Mathematics Summer Program.

Farewell to Tristan Miller

2023-12-15T00:00:00+01:00

Today we bid a fond farewell to Tristan Miller, who joined OFAI's research team in 2019. Tristan will be moving to Winnipeg to join the faculty of the University of Manitoba Department of Computer Science, where we wish him the best of success. He will continue his relationship with OFAI as an Associate Researcher.

MEi:CogSci students visit OFAI

2023-12-14T00:00:00+01:00

Today OFAI hosted a visit from students of the Middle European Interdisciplinary Master's Programme in Cognitive Science.

OFAI director Robert Trappl and researchers Paolo Petta, Friedrich Neubarth, and Peter Hallman met with students to discuss past and ongoing research projects.

Dr. Klaus M. Stiefel, Silliman University and Neurolinx Research Institute

2023-12-13T00:00:00+01:00

It seems clear that the development and deployment of an artificial superintelligence, or ASI, would require tremendous energy requirements. Whether this poses an insurmountable barrier to the emergence of artificial general intelligence is discussed in "The Energy Challenges of Artificial Superintelligence", a talk by Klaus M. Stiefel of Silliman University and Neurolinx Research Institute. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 13 December 2023 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: We argue here that contemporary semiconductor computing technology poses a significant if not insurmountable barrier to the emergence of any artificial general intelligence system, let alone one anticipated by many to be “superintelligent”. This limit on artificial superintelligence (ASI) emerges from the energy requirements of a system that would be more intelligent but orders of magnitude less efficient in energy use than human brains. An ASI would have to supersede not only a single brain but a large population given the effects of collective behavior on the advancement of societies, further multiplying the energy requirement. A hypothetical ASI would likely consume orders of magnitude more energy than what is available in highly-industrialized nations. We estimate the energy use of ASI with an equation we term the “Erasi equation”, for the Energy Requirement for Artificial SuperIntelligence. Additional efficiency consequences will emerge from the current unfocussed and scattered developmental trajectory of AI research. Taken together, these arguments suggest that the emergence of an ASI is highly unlikely in the foreseeable future based on current computer architectures, primarily due to energy constraints, with biomimicry or other new technologies being possible solutions.

Speaker biography: Dr. Klaus M. Stiefel did his undergraduate studies at the University of Vienna, and his doctoral work at the Max Planck Institute for Brain Research in Frankfurt, Germany. He then spent time at the Salk Institute in La Jolla, CA, USA, at OIST in Okinawa, at the University of Western Sydney in Australia and is presently based at Silliman University in the Philippines. Klaus is interested in fish ecology & biodiversity and in computational neuroscience.

Prof. Ivan Habernal, MSc, PhD, Paderborn University

2023-12-06T00:00:00+01:00

Natural language processing has brought us enormous benefits, but with it challenges and concerns surrounding privacy. These issues are tackled in "Privacy in Natural Language Processing: Are we There yet?", a talk by Ivan Habernal of Paderborn University. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk via Zoom on Wednesday, 6 December 2023 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: In this talk, I will explore the challenges and concerns surrounding privacy in natural language processing (NLP) and present potential solutions to address them. I will discuss the use of anonymization and differential privacy techniques to protect sensitive information while still enabling the training of accurate NLP models. Additionally, I will emphasize the importance of transparency and reproducibility when implementing privacy-preserving solutions in NLP.

Speaker biography: Ivan Habernal is an assistant professor for Natural Language Processing at Paderborn University, Germany, where he leads the Trustworthy Human Language Technologies group. His current research areas include privacy-preserving NLP, legal argument mining, and explainable and trustworthy models. His research track covers argument mining and computational argumentation, crowdsourcing, and serious games, among others.

Assoz. Prof. Mag. Dr. Dagmar Gromann, BSc, University of Vienna

2023-11-22T00:00:00+01:00

Conceptual metaphors present a powerful cognitive vehicle to transfer knowledge structures across domains. To what extent do today's pre-trained language models understand the operation of these metaphors? This question is addressed in "Do Large Language Models Grasp Metaphors?", a talk by Dagmar Gromann of the University of Vienna. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 22 November 2023 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Conceptual metaphors present a powerful cognitive vehicle to transfer knowledge structures from a source to a target domain, e.g. WORDS ARE WEAPONS as in "Your words pierce my heart". Prior neural approaches focus primarily on detecting whether natural language sequences are metaphoric or literal. In this talk, I will present work on probing metaphoric knowledge in pre-trained language models. The focus is on testing their capability to predict source domains given an input sentence and a target domain in English and Spanish. Several methods from fine-tuning to few-shot prompting are tested. Results show that the most common error type is the hallucination of source domains.

Speaker biography: Dagmar Gromann is Assistant Professor at the Centre for Translation Studies, University of Vienna, with prior post-doc positions at IIIA-CSIC in Barcelona, Spain and TU Dresden, Germany. Her research focuses on neural methods for knowledge extraction of cognitive and multilingual concepts as well as language technologies, including their socio-technical implications, e.g. gender bias. In terms of cognitive concepts, she is particularly interested in image schemas inspired by embodied cognition and conceptual metaphors. Furthermore, she has co-created a new master's program called Multilingual Technologies that is jointly organized by the University of Vienna and FH Campus Wien and represents the first computational linguistic program in Austria.

Tristan Miller speaks at JKU Linz

2023-11-21T00:00:00+01:00

OFAI's Tristan Miller made a research visit today to the Johannes Kepler University Linz Institute for Machine Learning. He delivered an invited talk, "Computational Analysis and Translation of Wordplay", as part of the Lecture Series Artificial Intelligence for first-year students in the Bachelor's in Artificial Intelligence programme.

Friedrich Neubarth invited to the "Fachtagung Spracherkennung" at the Austrian Parliament

2023-11-10T00:00:00+01:00

Friedrich Neubarth was invited to the panel discussion of the "Fachtagung Spracherkennung" at the Austrian Parliament. This workshop targeted employees of parliaments and similar institutions from German speaking countries (including Luxemburg) and focused on the topic of the potentials of and potential problems with Automatic Speech Recognition (ASR) as an assisting tool for generating protocols of assembly meetings in the context of fostering democracy and transparency.

For further information, click HERE

Dr. phil. Dr. tech. Dipl.-Ing. Erich Prem, MBA, University of Vienna

2023-11-08T00:00:00+01:00

The interest in the ethics of AI systems has grown significantly over the last few years. An overview of some of the key ethical issues for AI, including technical solutionism and totalitarian tendencies of AI-based norm enforcement, is presented in "Ethics of AI: Good AI Versus the Totalitarian Enforcement of Norms", a talk by Erich Prem of the University of Vienna. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 8 November 2023 at 18:30 CET (UTC+1):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: The interest in the ethics of AI systems has grown significantly over the last few years as evidenced by a growing literature on the topic and a mounting body of strategies, proposed regulations, standards, and technical approaches. In this talk, we provide an overview of some of the key ethical issues discussed for AI systems such as trolley problems or systems that talk back (ChatGPT). We review the related challenges as well as some of the proposed technical solutions such as model cards or rules for online discourse. The talk will focus on open issues and critically discuss technical solutionism and totalitarian tendencies of AI-based norm enforcement.

Speaker biography: Erich Prem is a researcher in the Philosophy of Technology (Ethics and Epistemology) at the University of Vienna and teaches Digital Humanism at the Technical University of Vienna. He is Director of eutema, a strategic technology consultancy in Vienna, Austria. Earlier, Erich was a researcher at the Austrian Research Institute for AI (OFAI) and the MIT AI Lab. He holds a PhD in Philosophy (Epistemology) and a PhD in Computer Science (Artificial Intelligence). His research work is often interdisciplinary with a focus on digital humanism, AI ethics, RTDI strategy, and digital innovation.

Prof. Simon Penny, University of California Irvine

2023-10-23T00:00:00+02:00

OFAI is pleased to present "Skill: Know-how, Artisanal Practices and 'Higher' Cognition", a talk by Simon Penny of the University of California Irvine. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Monday, 23 October 2023 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Skilled practitioners attest that in their experience of skilled practice, intelligence feels like it is happening in peripersonal space, at the fingertips, on the workbench. This paper begins from the premise that skilled embodied practices are intelligence - as much improvisation as hylomorphism (Ingold) - enacted amongst tools, materials and cognitive ecologies. As a lifelong practitioner, I seek to remain grounded in practice, while pursuing an interdisciplinary inquiry into the concept of skill, engaging philosophy, psychology, anthropology, cognitive science and neuroscience. The experience of skilled practices destabilises the (received) skill-intelligence binary, which is seen as a corollary of the mind-body binary. A dualist framework that distinguishes ‘higher' and ‘lower’ cognition and valorises abstraction, is not conducive to optimal discussion of skill. I will discuss the historical construction of this privileging of abstraction in philosophy and theorisation of cognition. A different framework will be suggested, drawing upon concepts of know-how (Ryle), the ‘performative idiom’ (Pickering), enactivism (Varela, Thompson, DiPaolo), pre-reflective awareness (Legrand), epistemic action (Kirsh), cognitive ecologies (Hutchins, Sutton). Arguments from neuroscience are then marshalled, focusing on phylogenetics and on proprioception, in order to build a non-dualist approach to neurophysiology, that provides a more balanced theoretical framework within which to discuss skill and/as cognition. If embodied practices are taken to constitute intelligence, this has ramifications for general conceptualisations of intelligence, and in turn, for rhetorics validating artificial intelligence, and claims made for interactive screen-based pedagogies.

Speaker biography: Simon Penny is an artist and theorist with a longstanding focus on emerging technologies, embodied and situated aspects of artistic practices, and critical analysis of computer culture. Much of his career has been at the intersection of engineering and art – he has developed custom immersive, sensor-based systems for embodied interaction. He published Making Sense: Cognition, Computing, Art and Embodiment in 2017 (MIT press) and directed A Body of Knowledge: Embodied Cognition and the Arts conference (2016). As part of his current book project Skill, he is working to build a non-dualistic approach to neurophysiology as a basis for a discussion of skill vis-a-vis intelligence. A current preoccupation is with ways emerging technologies constrain scientific and applied research.

Originally from Australia, Penny was Professor of Art and Robotics at Carnegie Mellon (1993-2000). He founded the Arts Computation Engineering (ACE) graduate program at the University of California Irvine, 2001-2012. He was Labex International Professor, University Paris8 and ENSAD in 2014, and visiting professor at Cognitive Systems and Interactive Media masters, University Pompeu Fabra, Barcelona, 2006-2013. Penny is professor of Electronic Art and Design (Dept of Art) at University of California, Irvine, with appointments in the Department of Music and in Informatics.

Taubman College group visits OFAI

2023-10-12T00:00:00+02:00

Today OFAI hosted a research visit from the Architecture and Artificial Intelligence Laboratory (AR2IL) of the University of Michigan's Taubman College for Architecture and Urban Planning.

OFAI researchers met with AR2IL director Matias del Campo and a large contingent of graduate students. Participants introduced their past and ongoing research projects and discussed the future use of AI in architectural and other creative domains.

Prof. Dr. Nafise Sadat Moosavi, BSc, MSc, University of Sheffield

2023-10-11T00:00:00+02:00

While models must perform reasoning to understand human language, this is mostly required for downstream applications and not as a standalone skill. The challenges of end-to-end reasoning in downstream applications, with a specific emphasis on end-to-end arithmetic reasoning, is the topic of "Challenges of End-to-End Reasoning in NLP", a talk by Nafise Sadat Moosavi of the University of Sheffield. The talk is part of OFAI's 2023 Fall Lecture Series.

Members of the public are cordially invited to attend the talk via Zoom on Wednesday, 11 October 2023 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: To understand human language, language models have to perform various reasoning skills, e.g., logical reasoning, commonsense reasoning, temporal reasoning, etc. There are multiple datasets for directly evaluating each of these reasoning skills. However, these reasoning skills are mostly required for downstream applications and not as standalone skills. For instance, a model may need to perform arithmetic reasoning for answering a question or to correctly summarize a table. However, it is not clear whether a model that performs well on a dataset that is designed to evaluate arithmetic reasoning would also improve the results on a QA dataset that requires arithmetic reasoning. As a result, we should pay special attention to developing end-to-end models for downstream applications that are also capable of performing various reasoning skills. This presentation focuses on the challenges of end-to-end reasoning in downstream applications, with a specific emphasis on end-to-end arithmetic reasoning.

Speaker biography: Dr. Nafise Sadat Moosavi is a Lecturer in Natural Language Processing at the Computer Science Department of the University of Sheffield. Before joining the University of Sheffield, she was a postdoctoral researcher at the Technical University of Darmstadt. She works on the limitation of language models to improve their fairness, reasoning, robustness, and efficiency. She co-founded and co-organizes SustaiNLP workshops and regularly serves as a senior area chair and area chair at *ACL conferences.

JOKER workshop at CLEF 2023

2023-10-05T00:00:00+02:00

Papers from the workshop "JOKER: Automatic Wordplay Analysis" have now been published in the proceedings of CLEF 2023, the 14th Conference and Labs of the Evaluation Forum.

The JOKER workshop, co-organized by OFAI's Tristan Miller, aims to bring translators and computer scientists together to develop data and evaluation metrics for the (semi-)automatic translation of humorous language. The workshop consists of a number of shared and unshared tasks, held in the spring of 2023, followed by a series of sessions at the CLEF 2023 conference in Thessaloniki from 18 to 21 September, 2023.

A peer-reviewed overview of the entire workshop appears in Experimental IR Meets Multilinguality, Multimodality, and Interaction, a volume in Springer's Lecture Notes in Computer Science series. Papers describing the individual shared tasks of the workshop, as well as the participating systems, are published in Working Notes of CLEF 2023, a volume in the CEUR Workshop Proceedings series.

Full bibliographic details of the OFAI-coauthored papers are as follows:

Liana Ermakova, Tristan Miller, Anne-Gwenn Bosser, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. Overview of JOKER – CLEF-2023 track on automatic wordplay analysis. In Avi Arampatzis, Evangelos Kanoulas, Theodora Tsikrika, Stefanos Vrochidis, Anastasia Giachanou, Dan Li, Mohammad Aliannejadi, Michalis Vlachos, Guglielmo Faggioli, and Nicola Ferro, editors, Experimental IR Meets Multilinguality, Multimodality, and Interaction: Proceedings of the Fourteenth International Conference of the CLEF Association (CLEF 2023), volume 14163 of Lecture Notes in Computer Science (ISSN 0302-9743), pages 397–415, Cham, September 2023. Springer. ISBN 978-3-031-42448-9. DOI: 10.1007/978-3-031-42448-9_26.
Liana Ermakova, Tristan Miller, Anne-Gwenn Bosser, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. Overview of JOKER 2023 Automatic Wordplay Analysis Task 1 – pun detection. In Mohammad Aliannejadi, Guglielmo Faggioli, Nicola Ferro, and Michalis Vlachos, editors, Working Notes of CLEF 2023 – Conference and Labs of the Evaluation Forum, volume 3497 of CEUR Workshop Proceedings (ISSN 1613-0073), pages 1785–1803, October 2023.
Liana Ermakova, Tristan Miller, Anne-Gwenn Bosser, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. Overview of JOKER 2023 Automatic Wordplay Analysis Task 2 – pun location and interpretation. In Mohammad Aliannejadi, Guglielmo Faggioli, Nicola Ferro, and Michalis Vlachos, editors, Working Notes of CLEF 2023 – Conference and Labs of the Evaluation Forum, volume 3497 of CEUR Workshop Proceedings (ISSN 1613-0073), pages 1804–1817, October 2023.
Liana Ermakova, Tristan Miller, Anne-Gwenn Bosser, Victor Manuel Palma Preciado, Grigori Sidorov, and Adam Jatowt. Overview of JOKER 2023 Automatic Wordplay Analysis Task 3 – pun translation. In Mohammad Aliannejadi, Guglielmo Faggioli, Nicola Ferro, and Michalis Vlachos, editors, Working Notes of CLEF 2023 – Conference and Labs of the Evaluation Forum, volume 3497 of CEUR Workshop Proceedings (ISSN 1613-0073), pages 1818–1827, October 2023.

Robert Trappl delivers ISA 2023 keynote

2023-09-28T00:00:00+02:00

OFAI director Robert Trappl gave a keynote address at the Internet Summit Austria 2023, a technical and trade conference organized by Internet Service Providers Austria (ISPA) on 28 September 2023.

Prof. Trappl's talk was featured in the event's second session, dedicated to transferring technical advancements to human advances. The keynote address was followed by a podium discussion featuring Prof. Trappl, Florian Aigner of TU Wien, Eva Eggeling of Fraunhofer, and Mic Hirschbrich of Apollo.ai.

OFAI 2023 Fall Lecture Series

2023-09-26T00:00:00+02:00

OFAI is delighted to announce its 2023 Fall Lecture Series, featuring an eclectic lineup of internal and external speakers.

The talks are intended to familiarize attendees with the latest research developments in AI and related fields, and to forge new connections with those working in other areas. The main theme of the current series is large language models.

Subscribe to our newsletter or our RSS feed, or bookmark this web page, to receive further details for the individual talks.

11 October 2023 at 18:30 CEST (UTC+2)

Nafise Sadat Moosavi (University of Sheffield)

Challenges of End-to-End Reasoning in NLP

To understand human language, language models have to perform various reasoning skills, e.g., logical reasoning, commonsense reasoning, temporal reasoning, etc. There are multiple datasets for directly evaluating each of these reasoning skills. However, these reasoning skills are mostly required for downstream applications and not as standalone skills. For instance, a model may need to perform arithmetic reasoning for answering a question or to correctly summarize a table. However, it is not clear whether a model that performs well on a dataset that is designed to evaluate arithmetic reasoning would also improve the results on a QA dataset that requires arithmetic reasoning. As a result, we should pay special attention to developing end-to-end models for downstream applications that are also capable of performing various reasoning skills. This presentation focuses on the challenges of end-to-end reasoning in downstream applications, with a specific emphasis on end-to-end arithmetic reasoning.

23 October 2023 at 18:30 CEST (UTC+2)

Simon Penny (University of California Irvine and Nottingham Trent University)

Skill: Know-how, Artisanal Practices and 'Higher' Cognition

Skilled practitioners attest that in their experience of skilled practice, intelligence feels like it is happening in peripersonal space, at the fingertips, on the workbench. This paper begins from the premise that skilled embodied practices are intelligence - as much improvisation as hylomorphism (Ingold) - enacted amongst tools, materials and cognitive ecologies. As a lifelong practitioner, I seek to remain grounded in practice, while pursuing an interdisciplinary inquiry into the concept of skill, engaging philosophy, psychology, anthropology, cognitive science and neuroscience. The experience of skilled practices destabilises the (received) skill-intelligence binary, which is seen as a corollary of the mind-body binary. A dualist framework that distinguishes ‘higher' and ‘lower’ cognition and valorises abstraction, is not conducive to optimal discussion of skill. I will discuss the historical construction of this privileging of abstraction in philosophy and theorisation of cognition. A different framework will be suggested, drawing upon concepts of know-how (Ryle), the ‘performative idiom’ (Pickering), enactivism (Varela, Thompson, DiPaolo), pre-reflective awareness (Legrand), epistemic action (Kirsh), cognitive ecologies (Hutchins, Sutton). Arguments from neuroscience are then marshalled, focusing on phylogenetics and on proprioception, in order to build a non-dualist approach to neurophysiology, that provides a more balanced theoretical framework within which to discuss skill and/as cognition. If embodied practices are taken to constitute intelligence, this has ramifications for general conceptualisations of intelligence, and in turn, for rhetorics validating artificial intelligence, and claims made for interactive screen-based pedagogies.

8 November 2023 at 18:30 CET (UTC+1)

Erich Prem (University of Vienna)

Ethics of AI: Good AI Versus the Totalitarian Enforcement of Norms

The interest in the ethics of AI systems has grown significantly over the last few years as evidenced by a growing literature on the topic and a mounting body of strategies, proposed regulations, standards, and technical approaches. In this talk, we provide an overview of some of the key ethical issues discussed for AI systems such as trolley problems or systems that talk back (ChatGPT). We review the related challenges as well as some of the proposed technical solutions such as model cards or rules for online discourse. The talk will focus on open issues and critically discuss technical solutionism and totalitarian tendencies of AI-based norm enforcement.

22 November 2023 at 18:30 CET (UTC+1)

Dagmar Gromann (University of Vienna)

Do Large Language Models Grasp Metaphors?

Conceptual metaphors present a powerful cognitive vehicle to transfer knowledge structures from a source to a target domain, e.g. WORDS ARE WEAPONS as in "Your words pierce my heart". Prior neural approaches focus primarily on detecting whether natural language sequences are metaphoric or literal. In this talk, I will present work on probing metaphoric knowledge in pre-trained language models. The focus is on testing their capability to predict source domains given an input sentence and a target domain in English and Spanish. Several methods from fine-tuning to few-shot prompting are tested. Results show that the most common error type is the hallucination of source domains.

6 December 2023 at 18:30 CET (UTC+1)

Ivan Habernal (Paderborn University)

Privacy in Natural Language Processing: Are we There yet?

In this talk, I will explore the challenges and concerns surrounding privacy in natural language processing (NLP) and present potential solutions to address them. I will discuss the use of anonymization and differential privacy techniques to protect sensitive information while still enabling the training of accurate NLP models. Additionally, I will emphasize the importance of transparency and reproducibility when implementing privacy-preserving solutions in NLP.

13 December 2023 at 18:30 CET (UTC+1)

Klaus M. Stiefel (Silliman University and Neurolinx Research Institute)

The Energy Challenges of Artificial Superintelligence

We argue here that contemporary semiconductor computing technology poses a significant if not insurmountable barrier to the emergence of any artificial general intelligence system, let alone one anticipated by many to be “superintelligent”. This limit on artificial superintelligence (ASI) emerges from the energy requirements of a system that would be more intelligent but orders of magnitude less efficient in energy use than human brains. An ASI would have to supersede not only a single brain but a large population given the effects of collective behavior on the advancement of societies, further multiplying the energy requirement. A hypothetical ASI would likely consume orders of magnitude more energy than what is available in highly-industrialized nations. We estimate the energy use of ASI with an equation we term the “Erasi equation”, for the Energy Requirement for Artificial SuperIntelligence. Additional efficiency consequences will emerge from the current unfocussed and scattered developmental trajectory of AI research. Taken together, these arguments suggest that the emergence of an ASI is highly unlikely in the foreseeable future based on current computer architectures, primarily due to energy constraints, with biomimicry or other new technologies being possible solutions.

20 December 2023 at 18:30 CET (UTC+1)

Thomas Graf (Stony Brook University)

Linguistics and Symbolic Computation in a World of Large Language Models

Language has always played a central role in artificial intelligence, yet AI researchers and linguists have rarely seen eye to eye on things, in particular the status of subsymbolic/neural approaches to language. After decades of debates, it looks like the subsymbolic approaches have finally emerged victorious. Not only are large language models (LLMs) succeeding in incredibly complex real-world tasks, subsymbolic models are also rapidly gaining traction in some areas of theoretical linguistics, e.g. lexical semantics. This raises the question: will symbolic linguistics be left in the dust, or is this actually an opportunity for meaningful synergy between symbolic and subsymbolic approaches?

17 January 2024 at 18:30 CET (UTC+1)

Stefanie Höhl (University of Vienna)

Social Rhythms and Biobehavioral Synchrony in Early Human Development

Caregiver–infant interactions are characterized by interpersonal rhythms at different timescales, from nursery rhymes and interactive games to daily routines. These rhythms make the social environment more predictable for young children and facilitate interpersonal biobehavioral synchrony with their caregivers. In adults, the brain rhythms of interaction partners entrain to communicative rhythms, including speech, supporting mutual comprehension and communication. I will present recent evidence that this is also the case in the infant brain, especially when babies are addressed directly by their caregiver through infant-directed speech in naturalistic interactions. Through using simultaneous measures of neural and physiological rhythms, e.g., dual-fNIRS and dual-ECG, from caregiver and infant during live face-to-face interactions, we can further deepen our understanding of early interactional dynamics and their reciprocal nature. I will present our recent research identifying factors supporting the establishment of caregiver–infant neural synchrony, such as affectionate touch and vocal turn-taking. I will further discuss the functional links and dissociations between caregiver–infant synchrony on the neural and physiological levels. I will outline potential implications of this work and point out important future directions.

How to attend: Attend in person (OFAI, Freyung 6/6/7, 1010 Vienna), or via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.

You can add this event to your calendar.

31 January 2024 at 18:30 CET (UTC+1)

Clemens Heitzinger (TU Wien)

Reinforcement Learning and its Application in Medicine and Large Language Models

Reinforcement learning has been instrumental in many advances in AI in recent years. The most publicized is certainly the development of ChatGPT and large language models (LLM) in general; the last and crucial training step of ChatGPT is reinforcement learning with human feedback (RLHF). Still, in order to fully solve learning problems, statements about the reliability of the results are necessary in addition to convergence results. For example, reliability and trustworthiness of AI systems is of utmost importance in medicine and other safety critical areas. In this talk, reinforcement-learning algorithms for training LLM and for calculating optimal treatments of sepsis patients are described. The questions of convergence to an optimal policy and of reliability are addressed by PAC (probably approximately correct) estimates and other approaches to policy evaluation.

How to attend: Attend in person (OFAI, Freyung 6/6/7, 1010 Vienna), or via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.

You can add this event to your calendar.

14 February 2024 at 18:30 CET (UTC+1)

Brigitte Krenn and Stephanie Gross (Austrian Research Institute for Artificial Intelligence)

Bias in Language Models Illustrated by the Example of Gender

Abstract TBA

How to attend: Attend in person (OFAI, Freyung 6/6/7, 1010 Vienna), or via Zoom (meeting ID: 842 8244 2460; passcode: 678868), or dial in by phone.

You can add this event to your calendar.

Brigitte Krenn interviewed on big data

2023-09-12T00:00:00+02:00

Austrian public radio (ORF Ö1) has broadcast another interview with OFAI's Brigitte Krenn, "Data Science: Eine Wissenschaft nur für Daten?". Journalist Mariann Unterluggauer speaks with Dr. Krenn about the emerging discipline of data science and the technical and ethical challenges it must address.

Tech press on Usenet revitalization

2023-08-31T00:00:00+02:00

OFAI's work on Usenet has been covered in articles in The Register and TechSpot:

"USENET, the OG social network, rises again like a text-only phoenix" by Liam Proven
"Usenet is still alive, and the Big-8 Board is managing things again" by Alfonso Maruccia

Since 2020, OFAI researcher Tristan Miller has been helping to preserve and revitalize Usenet, the Internet's oldest federated discussion network. He has been speaking regularly about this work at technical and academic events and last year co-authored a retrospective article on Usenet in the journal Internet Histories.

ORF Ö1: Brigitte Krenn on chatbots

2023-08-27T00:00:00+02:00

Brigitte Krenn was interviewed by Austrian public radio (ORF Ö1). The interview and accompanying online article discuss the past, present, and future of artificial intelligence, with a focus on historical and modern-day chatbots such as Joseph Weizenbaum's Eliza and OpenAI's ChatGPT.

SIGUL talk on Austrian speech synthesis

2023-08-18T00:00:00+02:00

On 18 August 2023 Lorenz Gutscher delivered a talk at SIGUL 2023, the 2nd Annual Meeting of the Special Interest Group on Under-resourced Languages.

The talk, "Neural Speech Synthesis for Austrian Dialects with Standard German Grapheme-to-Phoneme Conversion and Dialect Embeddings", was about dialect embeddings and how to use them to shift a speaker from standard to dialect or dialect to standard. An online demo of the technology is available to try out, and the full paper can be read on ResearchGate.

SIGUL 2023, held in Dublin from 18 to 20 August 2023, is a satellite workshop of Interspeech 2023.

MEi:CogSci internship on human–robot collaboration

2023-07-28T00:00:00+02:00

Arina Polyanskaya, a Master's student at the University of Vienna's MEI:CogSci program, has joined the OFAI for a summer internship.

Arina holds a Bachelor's degree in psychology from Florida International University (FIU) in Miami. At OFAI she is working under the supervision of Mag. Dr. Brigitte Krenn on a project that investigates the errors and miscommunications that occur during collaborative tasks between a human participant and a cobot in a mixed reality (XR) environment. In the project, Arina will extend her skills in qualitative analysis for human–robot interaction (HRI) working on annotating and analyzing recorded interaction videos. She will also gain experience in working with larger sets of data recorded from the XR environment, including working with databases and with tools and techniques from data science and machine learning.

Spanish-language chapter on interactive wordplay translation

2023-07-25T00:00:00+02:00

A chapter co-authored by Tristan Miller, "La interacción entre el hombre y la máquina en la traducción de juegos de palabras", has been published under the Creative Commons Attribution licence in the edited volume La traducción audiovisual a través de la traducción automática y la posedición: prácticas actuales y futuras.

The chapter is a Spanish-language translation of "Human–computer interaction in pun translation", which was published last year in the Routledge volume Using Technologies for Creative-Text Translation.

The chapter presents PunCAT, an interactive electronic tool for the translation of puns. Following the strategies known to be applied in pun translation, PunCAT automatically translates each sense of the pun separately; it then allows the user to explore the semantic fields of these translations in order to help construct a plausible target-language solution that maximizes the semantic correspondence to the original.

The chapter and research are joint work with Prof. Waltraud Kolb of the University of Vienna's Centre for Translation Studies. La traducción audiovisual a través de la traducción automática y la posedición is edited by Laura Mejías-Climent and Julio de los Reyes Lozano and is published by Comares.

JOKER Corpus paper at SIGIR 2023

2023-07-24T00:00:00+02:00

A paper describing the JOKER Corpus, a parallel corpus for multilingual wordplay recognition, has been published in the proceedings of SIGIR 2023, the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval.

The paper, "The JOKER Corpus: English–French Parallel Data for Multilingual Wordplay Recognition", was jointly authored by Liana Ermakova of the University of Western Brittany, Anne-Gwenn Bosser of the École Nationale d’Ingénieurs de Brest, Adam Jatowt of the University of Innsbruck, and Tristan Miller of OFAI.

The JOKER Corpus was produced for use in JOKER-2023, an upcoming workshop and shared task on Automatic Wordplay Analysis. JOKER-2023 will be held during the CLEF 2023 conference in Thessaloniki from 18 to 21 September, 2023.

SIGIR 2023 will be held from 23–27 July 2023 in Taipei.

Book chapter on artificial social agents

2023-07-10T00:00:00+02:00

"The Role of Multimodal Data for Modeling Communication in Artificial Social Agents", a chapter authored by Stephanie Gross and Brigitte Krenn has been published in the edited volume Handbook of Human-Machine Systems.

The chapter presents a selection of multimodal datasets collected at OFAI to study human communication behavior in task-oriented scenarios. It also discusses the implications of the authors' findings for modelling communicative behavior in artificial social agents.

The handbook is edited by Giancarlo Fortino, David Kaber, Andreas Nürnberger, and David Mendonça, and is published by Wiley.

AI panel at the International Society for Humor Studies Conference

2023-07-06T00:00:00+02:00

Tristan Miller co-organized and spoke at the Humor and Artificial Intelligence Panel at the 2023 International Society for Humor Studies Conference (ISHS).

The annual panel, organized together with Julia Rayz of Purdue University and Christian F. Hemplemann of Texas A&M University–Commerce, provides a forum for the interdisciplinary dissemination and discussion of research in computational humour. Tristan Miller's talk, "The JOKER Corpus", presented a parallel English–French data set of wordplay that can be used to train and evaluate humour-aware applications in information retrieval and artificial intelligence.

ISHS 2023 was held at Boston University from 3 July to 7 July 2023.

OFAI at ASAI's 2023 Networking Event

2023-07-05T00:00:00+02:00

OFAI participated today in a networking event organized by the Austrian Society for Artificial Intelligence (ASAI).

The event, intended to further connect the Austrian AI community, featured lightning talks and posters from 14 institutes, research groups, and other organizations. OFAI delegates, including Brigitte Krenn and Tristan Miller, were on hand to present the Institute's recent, ongoing, and upcoming research projects.

OFAI scientists speak at Internal Comms Forum

2023-06-20T00:00:00+02:00

OFAI research scientists Michael Pucher and Tristan Miller were featured discussion panelists at the 2023 Internal Comms Forum. The event, held on 20 June at the historic Haus der Ingenieure in Vienna, brought together an international audience of internal communication managers from a wide range of industries.

Under the theme "Bringing AI into Comms", the panelists explored the growing role of artificial intelligence in transforming internal communications practices. Joining Miller and Pucher on the panel were Thomas Lidy, Senior Director of AI & Data Science at Utopia Music, and Amelia Hernández Osorio, a digital workplace specialist at Wienerberger AG.

The Internal Comms Forum is a business conference featuring discussions, talks, and side events on the future of work, digital equality, internal and corporate communications, and how AI changes comms roles. It is organized by Advatera, a membership-based network for digital, marketing, and comms managers.

Peter Hallman speaks at Olomouc Linguistics Colloquium

2023-06-09T00:00:00+02:00

On 9 June 2023, Peter Hallman delivered a talk at the Olomouc Linguistics Colloquium.

His talk, "Few ≠ Not Many", argues that differences between 'few' and 'no' support more recent analyses of 'few' as a semantically simplex degree quantifier, while an analysis along the lines of Edward S. Klima's is essentially correct for 'no'.

The Olomouc Linguistics Colloquium is a general linguistics conference held biannually at Palacký University Olomouc in Czechia.

2023 Winter/Spring Lecture Series videos

2023-06-06T00:00:00+02:00

Video recordings from OFAI's 2023 Winter/Spring Lecture Series are now available in our YouTube channel. The full playlist is also embedded on this page below.

Subscribe now to keep apprised of future videos from the Austrian Research Institute for Artificial Intelligence!

Prof. Stacy Marsella, BA, MS, PhD, Northeastern University and University of Glasgow

2023-05-17T00:00:00+02:00

Computational models of human behavior are used in a wide range of artifacts. The synergy between psychology and the engineering of these artifacts is the subject of "Engineering the Impact of Emotion on Human Behavior", a talk by Stacy Marsella of Northeastern University and the University of Glasgow. The talk is part of OFAI's 2023 Winter/Spring Lecture Series.

Members of the public are cordially invited to attend the talk via Zoom on Wednesday, May 17th at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: Computational models of human behavior are used in a wide range of artifacts. At a large scale, social simulations are being used, for example, to explore people’s response to a natural disaster. At a medium-scale, models of human decision-makers are being used to study social technical systems such as the pharmaceutical drug supply networks. At the individual scale, work on human-robot and human-agent interaction seeks to facilitate interaction by giving artificial agents models of their human partners. At the extreme of modeling individual human behavior, virtual replicas of humans are being crafted, facsimiles of people that can engage people in face-to-face interactions using the same verbal and nonverbal behavior people use. The designs of these various models heavily leverage psychological theories and data. Psychology and the social sciences, in turn, use these computational artifacts as means to formulate, test, and explore theories about human behavior. In this talk, I will first give a brief overview of my group’s work in social simulation, social technical systems, HRI and virtual humans. Then I will exemplify the synergy between psychology and the engineering of these artifacts from the perspective of my group’s work on developing and applying computational models of emotion.

Speaker biography: Stacy Marsella is a professor at Northeastern University, USA, Khoury College of Computer Sciences with a joint appointment in psychology and at the University of Glasgow, UK, Centre for Social, Cognitive & Affective Neuroscience. Prior to joining Northeastern, he was a research professor in the Department of Computer Science at the University of Southern California and a research director at the Institute for Creative Technologies. Previously, he held positions at USC’s Information Sciences Institute and Bell Labs. Marsella’s multidisciplinary research is grounded in the computational modeling of human cognition, emotion, and social behavior, as well as the evaluation of those models. Beyond its relevance to understanding human behavior, the work has seen numerous applications, including health interventions, social skills training, and planning operations. His applied work includes frameworks for large-scale social simulations and a range of techniques and tools for creating virtual humans, facsimiles of people that can engage in face-to-face interactions.

MEi:CogSci student joins DANCR project

2023-05-12T00:00:00+02:00

Patrick Burns, a first-year Master's student at the University of Vienna's MEI:CogSci program, has joined the DANCR project under the supervision of Mag. Dr. Brigitte Krenn.

DANCR is an AI-based tool designed for the artistic community, particularly in contemporary dance, where improvisation plays a crucial role. Its purpose is to support individual artists in their dance research by expanding and refining their repertoire of improvisations. Patrick's internship contributes to the evaluation of the system in which he is part of a multidisciplinary team of professional dancers, philosophers, and computer scientists. The goal is to create an evaluation framework that can accurately assess the artistic potential of the DANCR system as an improvisation tool and gain insight into human-robot interaction in the creative domain.

Assoz. Prof. Mag. Hannes Fellner, MA, PhD, University of Vienna

2023-05-10T00:00:00+02:00

In parts of the Ancient Silk Road, the Sanskrit, Tocharian, and Saka languages were written in a little-understood script known as Tarim Brahmi. Recent efforts at making these languages and the Tarim Brahmi script digitally accessible is the subject of "Digital Advances on the Ancient Silk Road", a talk by Hannes Fellner of the University of Vienna. The talk is part of OFAI's 2023 Winter/Spring Lecture Series.

Members of the public are cordially invited to attend the talk in person (OFAI, Freyung 6/6/7, 1010 Vienna) or via Zoom on Wednesday, 10 May 2023 at 18:30 CEST (UTC+2):

URL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQWFnQT09
Meeting ID: 842 8244 2460
Passcode: 678868

You can add this event to your calendar.

Talk abstract: From the 2nd century CE on, communities and monasteries developed along the trade routes of the ancient Silk Road in and around the Tarim Basin in today’s Xinjiang Uyghur Autonomous Region in the People’s Republic of China. These were centres of writing, copying, translating, and transmitting texts similar to the monasteries in medieval Europe. The old Indo-European languages Sanskrit, Tocharian, and Saka written in a Central Asian variant of the Indian Brahmi script – Tarim Brahmi – were the major languages in use in the Tarim Basin in the first millennium CE. In contrast to the writing traditions in medieval Europe, the ones on this part of the Silk Road are not well understood, mainly due to the fragmentary status of texts. I this talk, I will address recent efforts of making these languages and the Tarim Brahmi script digitally accessible and operable for philological, palaeographic, and linguistic research in the framework of the FWF-START project “The characters that shaped the Silk Road – A database and digital palaeography of Tarim Brahmi”. In the project, the text witnesses are linked to their digital facsimiles on the character level using Transkribus. All data concerning the texts is combined in an XML database and published through a web application. This allows to determine which text was written by whom, when, where, and how in order to

trace the evolution of Tarim Brahmi and its adaptation to the different languages
reveal the relationship between script types, languages, and genres
categorize countless text fragments that are so far unidentified (regarding language, provenance, date, genre etc.)
potentially (re)combine scattered fragments belonging to the same manuscript leaf
and, of course, to better understand literacy and writing culture in the Tarim Basin

Speaker biography: Hannes A. Fellner studied linguistics at the University of Vienna and received his PhD from Harvard University in 2013. He was a postdoctoral researcher at the University of Vienna and assistant professor at Leiden University. Since 2018 he is the principal investigator of a START-project funded by Austrian Science Fund dedicated to the research of the Central Asian variants of the Indian Brahmi script. He is currently associate professor for historical linguistics and digital philology at the University of Vienna. He is a member of the Young Academy of the Austrian Academy of Sciences and the director of the Austrian Institute for Research on China and Southeast Asia. His research interests include Indo-European nominal morphology, historical and comparative linguistics and philology of the Indo-European languages of the ancient Silk Road, and theoretical approaches to language change.

Lorenz Gutscher speaks at Graz–Wien Speech Workshop

2023-05-10T00:00:00+02:00

OFAI's Lorenz Gutscher gave a talk at the 2nd Graz–Wien Speech Workshop about "Utilizing Dialect Embeddings for Text-to-Speech Synthesis of Austrian Varieties". The workshop, held on 3–4 May 2023, was organized by the Signal Processing and Speech Communication Laboratory at the Graz University of Technology.

OFAI hosts MEi:CogSci intern

2023-05-08T00:00:00+02:00

Klára Petrovicka joined OFAI in September 2022 to complete an internship as part of her MEi:CogSci studies at the University of Vienna. Under the supervision of Mag. Dr. Brigitte Krenn, the goal of the internship was to analyse machine learning approaches on processing text-based empathic utterances. In particular, they worked on reproducing EmoBERT model results; the code is publicly available. EmoBERT is a emotion-based variant of the BERT transformer model, which was able to learn emotion representations, and which was also fine-tuned to classify empathetic intents such as consoling and sympathising. After the initial analysis of EmoBERT's performance, including the replication of the published results, the model was then applied to another dataset and an error analysis was performed.

While the internship has now come to an end, the cooperation between Klara and OFAI to continue work on empathetic processing models continues.

Tristan Miller speaks at University of Konstanz

2023-05-04T00:00:00+02:00

OFAI's Tristan Miller made a research visit today to the University of Konstanz Department of Linguistics. He delivered an invited talk, "Computational Analysis and Translation of Wordplay", and met with faculty and graduate students.

Article on Arabic degree quantification

2023-04-26T00:00:00+02:00

"Equative Degree Quantification in Damascene Arabic", an article by Peter Hallman, has been published in Perspectives on Arabic Linguistics XXXIV.

The paper documents a degree-equative construction in contemporary Syrian Arabic, using data collected by elicitation from five native speakers from the city of Damascus.

Perspectives on Arabic Linguistics XXXIV, edited by Mahmoud Azaz, brings together eleven peer-reviewed articles from the Annual Symposium on Arabic Linguistics held at Tucson, Arizona in 2020 .

Paolo Petta lecture postponed

2023-04-26T00:00:00+02:00

Due to unforeseen circumstances, we have had to postpone Paolo Petta's talk, originally scheduled for 26 April. A new date will be announced later. We apologize for the inconvenience and look forward to welcoming you to the next scheduled lecture on 10 May.

Journal article on superlative displacement

2023-04-24T00:00:00+02:00

"Superlative displacement in 'sandwich' scanarios", an article by Peter Hallman, has been published in volume 31 of Natural Language Semantics.

The article seeks to reconcile the ‘movement’ account of the interpretation of superlative and comparative degree quantifiers with a class of apparent counterexamples. It provide an analysis of contexts in which the unexpected interaction of degree quantifiers with other terms in their scope is a side effect of quantification over situations inherent in the degree quantifier itself.

Natural Language Semantics is a journal devoted to semantics and its interfaces in grammar, especially syntax. It is published by Springer and edited by Amy Rose Deal.

Peter Hallman delivers deadjectivals workshop keynote

2023-04-21T00:00:00+02:00

Peter Hallman was a keynote speaker at the Workshop on Deadjectival Verb Formation in Indo-European, held on 10 and 11 March 2023 in Vienna at the Austrian Academy of Sciences's Institute of Iranian Studies.

The keynote talk, "A>V in the English Passive Participles", discusses the English passive participle morpheme as a case of historical adjective-to-verb derivation by virtue of loss of function of the original adjectivizing morphology.

Federation and Moderation at LibrePlanet 2023

2023-04-19T00:00:00+02:00

Usenet, the Internet's original federated discussion network, was the subject of a talk presented at the LibrePlanet 2023 conference. Together with fellow Big-8 Management Board member Rayner Lucas, OFAI's Tristan Miller introduced the technical underpinnings of the network and discussed its influence on and relevance to the modern Internet.

A full video of the talk, "Federation and Moderation: Usenet as the Original Decentralized Social Network", is now available for streaming on PeerTube and on MediaGoblin.

The LibrePlanet series of conferences is organized by the Free Software Foundation. LibrePlanet 2023 was held online and in Boston on 18 and 19 March 2023.