An Intelligent Conversational Agent Approach to Extracting Queries from Natural Language

to determine the answer the user's query. I. INTRODUCTION Typically, natural language database interfaces use parsing to identify and categorise the user's input [1]. The parsing process requires the creation of a lexicon, which includes both database specific and non-database specific terms. Different natural language database interfaces (NLDI's) use different methods to create the database specific terms. Microsoft English Query includes an authoring tool, so that database users can populate the lexicon with synonyms and alternative phrasings for database elements such as relation and attribute names [2]. Additionally, words which define the relationship between different database elements can be added (i.e. Customers buy products). Masque/SQL takes a similar approach, helping users to define the relationships between database elements using an " is-a " hierarchy [3]. English Language Frontend (ELF) [4] and Precise [5] find database specific terms by extracting information on the schema and entity names from a database, and using a dictionary to identify possible synonyms. In all these cases, once user input has been parsed, the relevant elements of user input are identified and related to the