A Natural Language Interface to InformationRetrieval

Increasing attention is being focused on Information Retrieval (IR) systems. Even as IR systems become more powerful, users of such systems encounter great diiculty in using them. This is because there is a conceptual gap between the users' information needs and the representations in IR systems. IR queries are notoriously complex. We propose a model-based solution to this problem. Our implemented system allows a user to pose queries in natural language. These queries are semantically processed to identify the concept the user is interested in. Heuristics are used to produce queries for the concept that the backend IR system can process. A single query by the user can be mapped to multiple information sources, including structured databases (where applicable). Our query generator naturally accommodates user and resource models to intelligently determine what information they seek.