Paper: Corpus Statistics Meet The Noun Compound: Some Empirical Results

ACL ID P95-1007
Title Corpus Statistics Meet The Noun Compound: Some Empirical Results
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 1995
Authors
  • Mark Lauer (Microsoft Institute, North Ryde Australia)

A variety of statistical methods for noun compound anMysis are implemented and compared. The results support two main conclusions. First, the use of conceptual association not only enables a broad cove- rage, but also improves the accuracy. Se- cond, an analysis model based on depen- dency grammar is substantially more accu- rate than one based on deepest constitu- ents, even though the latter is more preva- lent in the literature. 1 Background 1.1 Compound Nouns If parsing is taken to be the first step in taming the natural language understanding task, then broad co- verage NLP remains a jungle inhabited by wild be- asts. For instance, parsing noun compounds appears to require detailed world knowledge that is unavaila- ble outside a limited domain (Sparek Jones, 1983). Yet, far from being...