BEGIN:VCALENDAR
PRODID:-//Mozilla.org/NONSGML Mozilla Calendar V1.1//EN
VERSION:2.0
BEGIN:VTIMEZONE
TZID:Europe/Vienna
X-TZINFO:Europe/Vienna[2023c]
BEGIN:STANDARD
DTSTART:18930401T000000
RDATE:18930401T000000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+010521
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19161001T010000
RDATE:19161001T010000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19170917T030000
RRULE:FREQ=YEARLY;BYMONTH=9;BYDAY=3MO;UNTIL=19180916T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19200913T030000
RDATE:19200913T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19421102T030000
RDATE:19421102T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19431004T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1MO;UNTIL=19441002T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19450412T030000
RDATE:19450412T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19461007T030000
RDATE:19461007T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19471005T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU;UNTIL=19481003T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19800928T000000
RDATE:19800928T000000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19810927T030000
RRULE:FREQ=YEARLY;BYMONTH=9;BYDAY=-1SU;UNTIL=19950924T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19961027T030000
RDATE:19961027T030000
TZNAME:Europe/Vienna(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:STANDARD
DTSTART:19971026T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
TZNAME:(STD)
TZOFFSETFROM:+020000
TZOFFSETTO:+010000
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:19160430T230000
RDATE:19160430T230000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19170416T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=3MO;UNTIL=19180415T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19200405T020000
RDATE:19200405T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19400401T020000
RDATE:19400401T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19430329T020000
RDATE:19430329T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19440403T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1MO;UNTIL=19450402T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19460414T020000
RDATE:19460414T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19470406T020000
RDATE:19470406T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19480418T020000
RDATE:19480418T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19800406T000000
RDATE:19800406T000000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19810329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU;UNTIL=19960331T020000
TZNAME:Europe/Vienna(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:19970330T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
TZNAME:(DST)
TZOFFSETFROM:+010000
TZOFFSETTO:+020000
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
CREATED:20231006T115602Z
LAST-MODIFIED:20231010T111932Z
DTSTAMP:20231010T111932Z
UID:f8a9e219-cd4c-4a37-9593-e2c612da6ac6
SUMMARY:Talk by Thomas Graf: Linguistics and Symbolic Computation in a Worl
 d of Large Language Models
ORGANIZER;PARTSTAT=NEEDS-ACTION;CUTYPE=INDIVIDUAL:mailto:tristan.miller@ofa
 i.at
DTSTART;TZID=Europe/Vienna:20231220T183000
DTEND;TZID=Europe/Vienna:20231220T200000
DESCRIPTION;ALTREP="data:text/html,%3Cdiv%20class%3D%22col%20running-text%2
 2%3E%0A%20%20%20%20%20%20%20%20%20%20%3Cp%3EGiven%20the%20rapid%20rise%20of
 %20large%20language%20models%2C%20will%20symbolic%20linguistics%0A%20be%20l
 eft%20in%20the%20dust%2C%20or%20is%20this%20actually%20an%20opportunity%20f
 or%20meaningful%20%0Asynergy%20between%20symbolic%20and%20subsymbolic%20app
 roaches%3F%20%20This%20question%20is%20%0Aaddressed%20in%20%3Cstrong%3E%22L
 inguistics%20and%20Symbolic%20Computation%20in%20a%20World%20of%20Large%20L
 anguage%20Models%22%3C%2Fstrong%3E%2C%20a%20talk%20by%20%3Cstrong%3EThomas%
 20Graf%3C%2Fstrong%3E%20of%20%3Cstrong%3EStony%20Brook%20University%3C%2Fst
 rong%3E.%20%20The%20talk%20is%20part%20of%20%3Ca%20href%3D%22https%3A%2F%2F
 www.ofai.at%2Fevents%2Flectures2023fall.html%22%3EOFAI's%202023%20Fall%20Le
 cture%20Series%3C%2Fa%3E.%3C%2Fp%3E%0A%3Cp%3EMembers%20of%20the%20public%20
 are%20cordially%20invited%20to%20attend%20the%20talk%20via%20Zoom%20on%20%3
 Cstrong%3EWednesday%2C%2012%20December%202023%20at%2018%3A30%20CET%20(UTC%2
 B1)%3C%2Fstrong%3E%3A%3C%2Fp%3E%0A%3Cp%3EURL%3A%20%3Ca%20href%3D%22https%3A
 %2F%2Fus06web.zoom.us%2Fj%2F84282442460%3Fpwd%3DNHVhQnJXOVdZTWtNcWNRQllaQWF
 nQT09%22%3Ehttps%3A%2F%2Fus06web.zoom.us%2Fj%2F84282442460%3Fpwd%3DNHVhQnJX
 OVdZTWtNcWNRQllaQWFnQT09%3C%2Fa%3E%3Cbr%3E%0AMeeting%20ID%3A%20842%208244%2
 02460%3Cbr%3E%0APasscode%3A%20678868%3C%2Fp%3E%0A%0A%3Cp%3E%3Cstrong%3ETalk
 %20abstract%3A%3C%2Fstrong%3E%20Language%20has%20always%20played%20a%20cent
 ral%20role%20in%20artificial%20intelligence%2C%20yet%20AI%20researchers%20a
 nd%20linguists%20have%20rarely%20seen%20eye%20to%20eye%20on%20things%2C%20i
 n%20particular%20the%20status%20of%20subsymbolic%2Fneural%20approaches%20to
 %20language.%20After%20decades%20of%20debates%2C%20it%20looks%20like%20the%
 20subsymbolic%20approaches%20have%20finally%20emerged%20victorious.%20Not%2
 0only%20are%20large%20language%20models%20(LLMs)%20succeeding%20in%20incred
 ibly%20complex%20real-world%20tasks%2C%20subsymbolic%20models%20are%20also%
 20rapidly%20gaining%20traction%20in%20some%20areas%20of%20theoretical%20lin
 guistics%2C%20e.g.%C2%A0lexical%20semantics.%20This%20raises%20the%20questi
 on%3A%20will%20symbolic%20linguistics%20be%20left%20in%20the%20dust%2C%20or
 %20is%20this%20actually%20an%20opportunity%20for%20meaningful%20synergy%20b
 etween%20symbolic%20and%20subsymbolic%20approaches%3F%3Cbr%3E%3Cbr%3EIn%20t
 his%20talk%2C%20I%20argue%20for%20the%20latter%20by%20presenting%20%E2%80%9
 Csubregular%20syntax%E2%80%9D%20as%20a%20concrete%20example%20of%20what%20s
 uch%20a%20synergy%20may%20look%20like.%20Subregular%20syntax%20is%20a%20sym
 bolic%20approach%20that%20combines%20formal%20language%20theory%20with%20th
 e%20Minimalist%20syntax%20framework%20proposed%20by%20Noam%20Chomsky%2C%20w
 hich%20grants%20it%20a%20large%20degree%20of%20empirical%20coverage%20acros
 s%20a%20wide%20range%20of%20typologically%20diverse%20languages.%20Despite%
 20that%20broad%20coverage%2C%20subregular%20syntax%20is%20a%20very%20simple
 %20formalism%20that%20analyzes%20all%20syntactic%20dependencies%20in%20term
 s%20of%20relativized%20adjacency%20conditions.%20Even%20though%20these%20co
 nditions%20are%20stated%20over%20trees%2C%20they%20can%20actually%20be%20re
 duced%20to%20a%20very%20specific%20types%20of%20n-grams%20over%20strings.%2
 0This%20opens%20up%20a%20new%20way%20of%20representing%20sentence%20structu
 re%20in%20neural%20networks%20while%20bringing%20robust%20learning%20algori
 thms%20like%20stochastic%20gradient%20descent%20to%20Minimalist%20syntax.%2
 0It%20also%20casts%20doubt%20on%20claims%20in%20the%20literature%20that%20t
 he%20behavior%20of%20neural%20networks%20in%20specific%20linguistic%20tasks
 %2C%20e.g.%C2%A0binding%20or%20NPI-licensing%2C%20shows%20that%20they%20use
 %20tree%20structure.%20Instead%2C%20these%20findings%20may%20be%20indicativ
 e%20of%20a%20network%E2%80%99s%20ability%20to%20use%20fairly%20elaborate%20
 types%20of%20n-grams.%20The%20careful%20study%20of%20the%20symbolic%20appro
 ach%20of%20subregular%20syntax%20thus%20is%20an%20opportunity%20to%20deepen
 %20our%20understanding%20of%20neural%20networks%20while%20also%20harnessing
 %20their%20advantages%20for%20theoretical%20linguistics.%20%3C%2Fp%3E%0A%3C
 p%3E%3Cstrong%3ESpeaker%20biography%3A%20%3C%2Fstrong%3EThomas%20Graf%20is%
 20Associate%20Professor%20of%20Computational%20Linguistics%20in%20the%20Dep
 artment%20of%20Linguistics%20at%20Stony%20Brook%20University.%20He%20also%2
 0holds%20an%20affiliate%20appointment%20in%20Stony%20Brook%E2%80%99s%20Inst
 itute%20for%20Advanced%20Computational%20Science.%20Before%20joining%20Ston
 y%20Brook%2C%20he%20studied%20linguistics%20at%20the%20University%20of%20Vi
 enna%20and%20received%20his%20PhD%20from%20UCLA%20in%202013.%20His%20resear
 ch%20operates%20at%20the%20intersection%20of%20theoretical%20linguistics%2C
 %20computational%20linguistics%2C%20and%20cognitive%20science%2C%20with%20a
 %20particular%20focus%20on%20syntax%20(sentence%20structure).%20He%20is%20t
 he%20recipient%20of%20the%202014%20E.W.%20Beth%20Dissertation%20Prize%20for
 %20outstanding%20PhD%20theses%20in%20logic%2C%20language%2C%20and%20informa
 tion%2C%20and%20in%202019%20he%20received%20an%20NSF%20CAREER%20award%20for
 %20his%20research%20on%20subregular%20syntax.%20He%20is%20the%20creator%20o
 f%20the%20blog%20Outdex%2C%20which%20covers%20computational%20and%20theoret
 ical%20linguistics%2C%20and%20he%20loves%20to%20introduce%20high%20school%2
 0students%20to%20the%20wonders%20of%20computational%20linguistics%20during%
 20Stony%20Brook%E2%80%99s%20Mathematics%20Summer%20Program.%20%3C%2Fp%3E%0A
 %20%20%20%20%20%20%20%20%3C%2Fdiv%3E":\n          \n\nGiven the rapid rise 
 of large language models\, will symbolic linguistics\n be left in the dust\
 , or is this actually an opportunity for meaningful \nsynergy between symbo
 lic and subsymbolic approaches?  This question is \naddressed in "Linguisti
 cs and Symbolic Computation in a World of Large Language Models"\, a talk b
 y Thomas Graf of Stony Brook University.  The talk is part of OFAI's 2023 F
 all Lecture Series.\n\nMembers of the public are cordially invited to atten
 d the talk via Zoom on Wednesday\, 20 December 2023 at 18:30 CET (UTC+1):\n
 \nURL: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTWtNcWNRQllaQW
 FnQT09\n\nMeeting ID: 842 8244 2460\n\nPasscode: 678868\n\n\nTalk 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 pa
 rticular the status of subsymbolic/neural approaches to language. After dec
 ades of debates\, it looks like the subsymbolic approaches have finally eme
 rged victorious. Not only are large language models (LLMs) succeeding in in
 credibly complex real-world tasks\, subsymbolic models are also rapidly gai
 ning traction in some areas of theoretical linguistics\, e.g. lexical seman
 tics. This raises the question: will symbolic linguistics be left in the du
 st\, or is this actually an opportunity for meaningful synergy between symb
 olic and subsymbolic approaches?\n\nIn this talk\, I argue for the latter b
 y presenting “subregular syntax” as a concrete example of what such a s
 ynergy may look like. Subregular syntax is a symbolic approach that combine
 s formal language theory with the Minimalist syntax framework proposed by N
 oam 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 the
 se conditions are stated over trees\, they can actually be reduced to a ver
 y specific types of n-grams over strings. This opens up a new way of repres
 enting 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 network
 s in specific linguistic tasks\, e.g. binding or NPI-licensing\, shows that
  they use tree structure. Instead\, these findings may be indicative of a n
 etwork’s ability to use fairly elaborate types of n-grams. The careful st
 udy of the symbolic approach of subregular syntax thus is an opportunity to
  deepen our understanding of neural networks while also harnessing their ad
 vantages for theoretical linguistics. \n\nSpeaker biography: Thomas Graf is
  Associate Professor of Computational Linguistics in the Department of Ling
 uistics at Stony Brook University. He also holds an affiliate appointment i
 n Stony Brook’s Institute for Advanced Computational Science. Before join
 ing Stony Brook\, he studied linguistics at the University of Vienna and re
 ceived his PhD from UCLA in 2013. His research operates at the intersection
  of theoretical linguistics\, computational linguistics\, and cognitive sci
 ence\, with a particular focus on syntax (sentence structure). He is the re
 cipient of the 2014 E.W. Beth Dissertation Prize for outstanding PhD theses
  in logic\, language\, and information\, and in 2019 he received an NSF CAR
 EER award for his research on subregular syntax. He is the creator of the b
 log Outdex\, which covers computational and theoretical linguistics\, and h
 e loves to introduce high school students to the wonders of computational l
 inguistics during Stony Brook’s Mathematics Summer Program. \n\n        
LOCATION:Via Zoom: https://us06web.zoom.us/j/84282442460?pwd=NHVhQnJXOVdZTW
 tNcWNRQllaQWFnQT09
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR