Natural Language Understanding, Semantic-based Information Retrieval and Knowledge Management

Junichi Tsujii

Abstract:Natural language understanding (NLU) has been taken as one of the core fields, and the hardest one among others, of Artificial Intelligence. As such, it has been considered an impractical technology from the perspective of the real world application. However, the situation surrounding NLU has been changing rapidly. We have witnessed large knowledge resources being constructed, manually or automatically from huge collection of text and structured data bases in Web. We have also witnessed significant amount observation data of user behaviors being accumulated in the forms of query logs, query-click logs, networks among users, etc. These observation data will shed (and actually are shedding) light on how linguistic expressions and their interpretation interact in context. What was considered impractical as research scenario five years ago is now being taken as serious research and application scenario. In this talk, I will discuss how the three fields of natural language understanding, semantic-based information retrieval and knowledge management are being evolved into a single unified research field and how semantics of text and intentions manifested in user queries are to be computationally manipulated.