Open seminar:Question Classification
Date: June 16, 2005 (Thursday) at 14:00
Speaker: Empar Bisbal Asensi
In the last years, Question Answering (QA) has become one of the main challenges in Natural Language Processing (NLP) research area. It tries to obtain exact answers to questions formulated in natural language. That's the main difference between QA and Information Retrieval (IR) systems, which return a list of documents that may contain the answer somewhere in.
Question Classification (QC) is usually the first stage in a question answering system. Most QC systems are based on the use of heuristic rules and manually defined patterns. These systems are strongly domain dependent and they need a great amount of human work to formulate the patterns.
QC modules based on machine learning arose with the idea of building flexible applications. We have developed a multilingual SVM-based question classification system, with the main purpose of being language and domain independent. Due to this main goal it uses only surface text features. The system's performance has been tested on the questions of the TREC QA track and it seems to be promising.
Last modified Dec 9, 2011 12:59 pm