Cascaded finite-state partial parsing: A larger-first approach

A larger-first approach to partial parsing is presented. This approach is opposite to current easy-first approaches, and is capable of producing a more detailed partial parse. Syntactic relations are identified in descending order of their size. Special consideration is given to the comma punctuation mark. Its presence is used in identifying large syntactic relations and establishing a boundary for containment of ambiguity. The system is tested on the Wall Street Journal section of the Penn Treebank III and compared against a well-known easy-first approach.