The problem of determining the correct attachment sites for PPs when parsing natural language is investigated. Semantic information is introduced into the parsing process by grouping lexical items according to the semantic associations given in WordNet V5. 1. Data extracted from a partially parsed corpus is tagged with the semantic groupings. The resultant attachment preference patterns displayed in the data are used to assist the parsing.
|Title of host publication||Abstracts for the ANLP Post Graduate Workshop|
|Number of pages||8|
|Publication status||Published - 1998|