Abstract:Conversion between different grammar frameworks is of great importance to comparative performance analysis of the parsers developed based on them and to discover the essential nature of languages. This paper presents an approach that converts Combinatory Categorial Grammar (CCG) derivations to Penn Treebank (PTB) trees using a maximum entropy model. Compared with previous work, the presented technique makes the conversion practical by eliminating the need to develop mapping rules manually and achieves state-of-the-art results.

A Machine Learning Approach to Convert CCGbank to Penn Treebank.