25.02.2020: Paola Merlo

Mechanisms of argument structure realisation in human languages

 

Recent enlightening  overviews on the  emergence of human  syntax have proposed  that  "syntax is  tightly  linked  with  how events  in  the external world are perceived, structured and mentally represented."

The valency of a verb is  expressed through its arguments in very many ways.  Arguments can be explicit or implicit, they vary in the mapping of  their  semantic  role  onto their  grammatical  function  in  verb alternations,  they  vary in  their  way  of marking  the  grammatical function, by word order or case, among other ways.

Current  large-scale   treebank  annotation  projects   have  produced naturalistic data  annotated with  relevant information for  very many languages.   These  treebanks  enable  the  precise  and  fine-grained quantification of argument structure variability at the sentence level for many languages and support computational models of its learning.

In  this  talk,  I  will  provide an  overview  of  past  and  current computational approaches  to verb lexical semantics  which provide an articulate view  of the  mapping from event  structure to  syntax.  In particular I will illustrate cases of mismatches of argument structure and grammatical  functions (subjects that  are not agents  and objects that are not patients).  Looking at these minority mappings and how to automatically learn  them in  simple and complex  computational models could  provide  crucial  minimal   experimental  pairs  to  study  the evolutionary emerge of syntax.

Bio

Paola Merlo is associate  professor in the Linguistics department of the University of Geneva.  She is the head of the interdisciplinary research  group Computational  Learning and  Computational Linguistics (CLCL).   The  group  is  concerned  with  interdisciplinary  research combining  linguistic  modelling  with  machine  learning  techniques. Prof.  Merlo has  been editor of Computational Linguistics,  and is an ACL  Fellow.  Prof.  Merlo  holds  a  doctorate   in  Computational Linguistics  from  the University  of  Maryland,  USA.  She  has  been associate  research  fellow at  the  University  of Pennsylvania,  and visiting scholar at Rutgers, Edinburgh, Stanford and Uppsala.