Paper: Tree Topological Features for Unlexicalized Parsing

ACL ID C10-2014
Title Tree Topological Features for Unlexicalized Parsing
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010

As unlexicalized parsing lacks word to- ken information, it is important to inves- tigate novel parsing features to improve the accuracy. This paper studies a set of tree topological (TT) features. They quantitatively describe the tree shape dominated by each non-terminal node. The features are useful in capturing lin- guistic notions such as grammatical weight and syntactic branching, which are factors important to syntactic proc- essing but overlooked in the parsing lit- erature. By using an ensemble classifier- based model, TT features can signifi- cantly improve the parsing accuracy of our unlexicalized parser. Further, the ease of estimating TT feature values makes them easy to be incorporated into virtually any mainstream parsers.