Natural Language Interfaces to Domain Specific Knowledge Bases: An Illustration for Querying Elements of the Periodic Table

Abstract

The task of providing Natural Language Interface (NLI) to any domain specific knowledge base is much demanding despite (potentially) favorable factors like low volume of vocabulary, unambiguous and precise meaning of words, less number of relations among the entities, etc. The simplification of this task has been proposed and presented in this research. The authors have made a successful effort to develop an NLI system to answer the user's simple queries (in English) about the properties of chemical elements and their grouping in the Periodic Table. Adding to the ease, the user is not required to know anything about the structure of the knowledge base of the elements, since the software is implemented (using Logic Programming constructs) in Prolog wherein program and data are treated indistinguishably. Firstly, the system accepts a query and subsequently, it can analyze and understand the query, if the query contains all words within the domain specific vocabulary. Finally, it efficiently searches the knowledge base to answer the query, by reducing search space using artificial intelligence techniques (like symbolic manipulation). If the query is not understood by the system, it reports to the user the words not available in the knowledge base and the particular relations among the entities which could not be set. The knowledge base (~150 KB) contains the properties of chemical elements, their arrangement in Periodic Table and the inter-relationships among these properties. In a nutshell, the research suggests that to develop an NLI to a domain specific knowledge base, it is better to develop a parser capable of handling the entities and their interrelationships as understood in the domain; hence, only little is to be coded for the various grammars, languages, transition networks, etc