Extracting Syntax Statistics from Large Corpora of Written English

The field of linguistics has seen a growing interest in the statistics of everyday language. In studying how we acquire language and why some of its aspects are more difficult for us than others, it is critical to understand the linguistic environment to which we are exposed. However, gathering statistics over syntactic structures, even with a syntactically tagged corpus, can be difficult and time consuming. This report describes a partially automated method that alleviates many of the problems associated with gathering syntax statistics from parsed corpora. The method is then used to analyze a variety of structures in the Wall Street Journal and Brown corpora of written English. These structures include verb phrases, relative clauses, sentential noun phrases, prepositional phrases, and coordinate and subordinate clauses.