Paper: Learning Structured Classifiers for Statistical Dependency Parsing

ACL ID N07-3002
Title Learning Structured Classifiers for Statistical Dependency Parsing
Venue Human Language Technologies
Session Doctoral Consortium
Year 2007
Authors

My research is focused on developing ma- chine learning algorithms for inferring de- pendency parsers from language data. By investigating several approaches I have developed a unifying perspective that al- lows me to share advances between both probabilistic and non-probabilistic meth- ods. First, I describe a generative tech- nique that uses a strictly lexicalised pars- ing model, where all the parameters are based on words and do not use any part- of-speech (POS) tags nor grammatical cat- egories. Then, I incorporate two ideas from probabilistic parsing word similar- ity smoothing and local estimation to improve the large margin approach. Fi- nally, I present a simpler and more ef- cient approach to training dependency parsers by applying a boosting-like proce- dure to standard training...