A Maximum Entropy Approach to
Question Answering
Irene Cramer
The Maximum Entropy model has
proved to be useful in a large range of different NLP
applications as for example POS-tagging, Named Entity
Recognition, or Machine Translation. In my talk I will
present a Maximum Entropy approach to Question Answering. To build such
model I derived a feature set from the Knowledge Master corpus
and computed the model parameters by means of the maxent toolkit
by Zhang Le. I will compare my approach with the results of
Ravichandran, Hovy, and Och (2003) who built up a baseline QA system
with only four features.
Ravichandran, Hovy, and Och (2003): Statistical QA - Classifier vs.
Re-ranker: What's the difference? In: Proceedings of the ACL Workshop
on Multilingual Summarization and Question Answering - Machine
Learning and Beyond, Sapporo, Japan.