Gabriele Antonio Musillo

Unlexicalised Hidden Variable Models of Split Dependency Grammars

This talk presents transforms of split dependency grammars into unlexicalised context-free grammars annotated with hidden symbols, that make learning and decoding algorithms for probabilistic context-free grammars available to projective dependency grammars. The Inside-Outside algorithm is used to estimate the hidden variables decorating such transforms. Our best unlexicalised grammar achieves an accuracy of 88% on the Penn Treebank data set, that represents a 50% reduction in error over previously published results on unlexicalised dependency parsing. Such performance shows that our unlexicalised models are able to capture both lexical and structural information that is relevant to parsing accuracy and suggests that we should reassess the relevance of massive lexicalisation to dependency parsing.

Official inquiries about AIIS should be directed to Alexandre Klementiev (klementi AT uiuc DOT edu)
Last update: 01/22/2008