A Review on Natural Language Interface for Database

摘要

Information is retrieved from database by query languages like SQL. But every user is not capable of writing SQL queries as they may unaware of database schema or structure. This has led to the development of Intelligent DataBase Systems (IDBS). That is, a natural language interface is provided for a database which enables non-expert users to query the database in natural language instead of having to go through the pain of mastering a query language. As a result, many intelligent Natural Language Interfaces for Database (NLIDB) have been developed, which provide flexible options for easy manipulation of queries. The NLIDB systems are expected to formulate a database query based on the request given in natural language. This paper gives a review on various NLIDB architectures and approaches used for developing NLIDB. Based on the study conducted, it is observed that almost every system reviewed fails to generate an SQL query accurately. Among the reviewed systems the PCFG based system is the most efficient one as they generate grammar trees and the one with maximum probability forms the SQL query. WordNet based system comes next as they make use of ontology. A minor flaw of the system is that the user request needs to be converted into an intermediate language before formulating the SQL query. So a new system that combines the features of these two systems can be devised so as to overcome the problems faced by the NLIDB systems.