Maximum Entropy Models
From HLT@INESC-ID
Date
- 15:00, Friday, December 4th, 2009
- Room 4
Speaker
Abstract
Maximum entropy models are discriminative models that have been recently applied to many NLP tasks, such as part-of-speech tagging, sentence boundary detection, NER. This approach has strong mathematical foundations and forms the core of more complicated, structured classification models, such as CRFs and MEMMs. This presentation shows the principle of maximum entropy and describes its application to different NLP tasks.