BEGIN:VCALENDAR
PRODID;X-RICAL-TZSOURCE=TZINFO:-//com.denhaven2/NONSGML ri_cal gem//EN
CALSCALE:GREGORIAN
VERSION:2.0
BEGIN:VEVENT
DTEND;VALUE=DATE-TIME:20200523T140000Z
DTSTART;VALUE=DATE-TIME:20200523T133000Z
DTSTAMP;VALUE=DATE-TIME:20200522T115533Z
UID:07de3fa9-b45b-4dd2-86ef-e38298a1324b@talks.stuts.de
DESCRIPTION:Why do natural languages develop morphological structure? Is 
 human cognition biased toward a lexicon that is morphologically structur
 ed? And if so\, how easily does a learner achieve morphological generali
 zation\, that is\, how easily may a paradigm be established in a system 
 given some set of cognitive constraints?\nQuestions such as these can be
  fruitfully addressed by using computational methods\, which provide a h
 ighly parametrized test ground for any models of cognition. Importantly\
 , in this case\, they allow a great degree of control over crucial confo
 unding factors\, such as\, for instance\, structural bias led by the spe
 cific typology of the native language of human subjects in experimental 
 conditions.\nIn this ongoing project\, led by Denis Paperno and Yoad Win
 ter at Utrecht University\, we implement a modified skipgram model using
  FastText word embeddings. Word embeddings allow us to represent both fo
 rm and meaning using vectors\, which are assigned to text words with res
 pect to their distributional properties. In the case of this model\, the
  distributional properties of subword n-grams are take into account to g
 enerate the pre-trained word vectors we will use. The goal of this proje
 ct is to use iterations in order to test the extent to which an ideal le
 arner is able to generalize a morphological paradigm from a limited set 
 of examples in context\, thus extracting systematic morphological relati
 ons between strings in the input. We plan to use vector cosine similarit
 y\, among other test methods\, to assess how close together words of a s
 ame paradigm are represented. That is\, we test for ease of paradigm dis
 covery in an ideal learner\, assuming only distributional probabilities 
 of morphemes in the input.\nThe project is in its early stages\, and in 
 this talk we present its methods\, advancements\, expected results\, and
  current challenges.
URL:https://talks.stuts.de/de/stuts67/public/events/282
SUMMARY:Computational approaches to morphological relatedness: a pilot st
 udy on paradigm discovery
ORGANIZER:stuts67
LOCATION:stuts67 - De Saussure
END:VEVENT
END:VCALENDAR