SEEKER: a conversational agent as a natural language interface to a relational database

Smith, EV, Crockett, K ORCID: https://orcid.org/0000-0003-1941-6201, Latham, A ORCID: https://orcid.org/0000-0002-8410-7950 and Buckingham, F (2014) SEEKER: a conversational agent as a natural language interface to a relational database. In: World Congress on Engineering (WEC) 2014, 02 July 2014 - 04 July 2014, London, U.K..

Preview

Published Version Available under License In Copyright. Download (607kB) | Preview

Abstract

Managers of companies are typically not SQL (Structured Query Language) experts but require information 24/7. Therefore, a growing need for Natural Language Interfaces to Databases (NLIDs) has been identified, with a vast amount of research being undertaken in the area. The existing approaches to NLIDs present many weaknesses including the inability to deal with grammatical mistakes in user input, the inability to communicate with the user to correct mistakes and the inability to allow refinement of query results. This paper proposes a system, SEEKER, which uses a Conversational Agent (CA) as the Natural Language Interface (NLI) in a NLID. The CA is used to capture key words in the user's utterance. Once these key words have been identified, the most appropriate SQL template is selected by the expert system using rule based reasoning. The identified variables are mapped to the SQL template in order to create an SQL query. SEEKER allows for refinement of query results. SEEKER was evaluated in terms of user satisfaction and task completion. The results of the evaluation were promising.