BEGIN:VCALENDAR VERSION:2.0 PRODID:ILLC Website X-WR-TIMEZONE:Europe/Amsterdam BEGIN:VTIMEZONE TZID:Europe/Amsterdam X-LIC-LOCATION:Europe/Amsterdam BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:19700329T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:19701025T030000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:/NewsandEvents/Archives/2019/newsitem/10678/12 -March-2019-Computational-Linguistics-Seminar-Vict oria-Yaneva DTSTAMP:20190307T164535 SUMMARY:Computational Linguistics Seminar, Victori a Yaneva ATTENDEE;ROLE=Speaker:Victoria Yaneva (University of Wolverhampton) DTSTART;TZID=Europe/Amsterdam:20190312T160000 LOCATION:ILLC Seminar Room F1.15, Science Park 107 , Amsterdam DESCRIPTION:p> When processing a text, both humans and machines must cope with ambiguity and non-com positionality. These phenomena represent a conside rable challenge for NLP systems, while at the same time there is limited evidence from online measur es on how humans solve them during natural reading . We approach these two problems as one and hypoth esize that obtaining information on how humans pro cess ambiguous and non-compositional phrases can i mprove the computational treatment of such instanc es. I will present experiments on using eye-tracki ng data to improve NLP models for two tasks: class ifying the different roles of the pronoun It (nomi nal anaphoric, clause anaphoric and non-referentia l), as well as the identification of multi-word ex pressions. The experiments test whether gaze-based features improve the performance of state-of-the- art NLP models and the extent to which gaze featur es can be used to partially or entirely substitute the crafting of linguistic ones. The best-perform ing models are then analysed to better understand the cognitive processing of these linguistic pheno mena and findings are discussed with respect to th e E-Z model of reading and the processing stages d uring which disambiguation occurs. X-ALT-DESC;FMTTYPE=text/html:\n

p> When pro cessing a text, both humans and machines must cope with ambiguity and non-compositionality. These ph enomena represent a considerable challenge for NLP systems, while at the same time there is limited evidence from online measures on how humans solve them during natural reading. We approach these two problems as one and hypothesize that obtaining in formation on how humans process ambiguous and non- compositional phrases can improve the computationa l treatment of such instances. I will present expe riments on using eye-tracking data to improve NLP models for two tasks: classifying the different ro les of the pronoun It (nominal anaphoric, clause a naphoric and non-referential), as well as the iden tification of multi-word expressions. The experime nts test whether gaze-based features improve the p erformance of state-of-the-art NLP models and the extent to which gaze features can be used to parti ally or entirely substitute the crafting of lingui stic ones. The best-performing models are then ana lysed to better understand the cognitive processin g of these linguistic phenomena and findings are d iscussed with respect to the E-Z model of reading and the processing stages during which disambiguat ion occurs.

\n URL:http://projects.illc.uva.nl/LaCo/CLS/ END:VEVENT END:VCALENDAR