Linguistik online 17, 5/03

Learning to Disambiguate Syntactic Relations

Gerold Schneider (Zurich/Geneva)

 

Abstract

Natural Language is highly ambiguous, on every level. This article describes a fast broad-coverage state-of-the-art parser that uses a carefully hand-written grammar and probability-based machine learning approaches on the syntactic level. It is shown in detail which statistical learning models based on Maximum-Likelihood Estimation (MLE) can support a highly developed linguistic grammar in the disambiguation process.


 

full text