An Intelligent Conversational Agent Approach to Extracting Queries from Natural Language
Karen Pudner, Keeley A. Crockett, Zuhair A. Bandar · 2007
Abstract—This paper is concerned with the application of a conversational agent and expert system to provide a natural language interface to a database. Typically, natural language database interfaces (NLDI's) use grammatical and/or statistical parsing. Conversational agents take a different approach, capturing key elements of user input which then trigger pre-determined output templates. It is assumed that the type of natural language questions which could be asked of a specific relational database will contain a limited number of key words (attributes), which could be captured by a conversational agent. In the proposed system, once a conversational agent has identified all relevant attributes and their values, an expert system would then apply rule based reasoning on these attributes to construct an SQL query. The knowledge base of the expert system would contain information on the database structure (metadata) and on the different possible structures of SQL queries. This would result in a real time system, which could extract both database attributes and attribute values from the user input and automatically apply a rule based reasoning system to determine the answer the user’s query. Index Terms—Conversational Agents, Natural Language, SQL I.