Improving the Precision of a Closed-Domain Question-Answering System with Semantic Information

This paper presents our experiments in applying semantic information to improve the precision of the information retrieval module in a closed-domain question-answering system. That system aims at replying questions on services offered by a large company, here Bell Canada. Our approach consists in finding a set of special terms and building an ontological concept hierarchy, which can effectively characterize the relevance of a retrieved candidate to its corresponding question. Combining these two kinds of semantic information with the information retrieval module has resulted in very good improvements.

• Type: Conference
• Year: 2004
• Where: RIAO
• Authors: Hai Doan-Nguyen, Leila Kosseim
• Keywords: Closed-domain Question-answering System, Information Retrieval Module, RIAO 2004, RIAO 2007, Semantic Information