/ 0
60%
Table of contents

Document in text mode:

Proceedingsofthe2021ConferenceoftheNorthAmericanChapteroftheAssociationforComputationalLinguistics:HumanLanguageTechnologies,pages3091–3096June6–11,2021.©2021AssociationforComputationalLinguistics3091YouSoundLikeSomeoneWhoWatchesDramaMovies:TowardsPredictingMoviePreferencesfromConversationalInteractionsSergeyVolokhin*,JoyceHo*,OlegRokhlenko+,andEugeneAgichtein*+*EmoryUniversity,GAUSA+Amazon,Seattle,WA,USA{svolokh,joyce.c.ho}@emory.edu,{olegro,eugeneag}@amazon.comAbstractTheincreasingpopularityofvoice-basedper-sonalassistantsprovidesnewopportunitiesforconversationalrecommendation.Onepartic-ularlyinterestingareaismovierecommenda-tion,whichcanbenefitfromanopen-endedin-teractionwiththeuser,throughanaturalcon-versation.Weexploreonepromisingdirec-tionforconversationalrecommendation:map-pingaconversationaluser,forwhomthereislimitedornodataavailable,tomostsimi-larexternalreviewers,whosepreferencesareknown,byrepresentingtheconversationasauser’sinterestvector,andadaptingcollabora-tivefilteringtech...