Context-free Approximation of LTAG towards CFG Filtering

We present a method to approximate a LTAG grammar by a CFG. A key process in the approximation method is finite enumeration of partial parse results that can be generated during parsing. We applied our method to the XTAG English grammar and LTAG grammars which are extracted from the Penn Treebank, and investigated characteristics of the obtained CFGs. We perform CFG filtering for LTAG by the obtained CFG. In the experiments, we describe that the obtained CFG is useful for CFG filtering for LTAG parser.