The  Weizmann  Institute  of  Science
                  Faculty of Mathematics and Computer Science


                            Computer Science Seminar

                                   Mats Rooth
                            University of Stuttgart


                                 will speak on


               EM Induction of Computational Linguistic Lexicons

Abstract:
The talk will present research on learning lexicons for natural languages from
large text samples, using a methodology which combines symbolic grammars and
parsing with numerical optimization.  Inside-outside estimation with a
lexicalized probabilistic grammar is used to collect frequency observations of
words in their local syntactic context. In a subsequent step, two-dimensional
clusters of words which capture aspects of lexical meaning are induced using
the EM algorithm. These clusters are used to label argument positions of
verbs.





                      The lecture will take place in the 
                     Lecture Hall, Room 1, Ziskind Building
                           on Monday, January 3, 2000
                                    at 14:30