Natural Language Interfaces to Databases Meikiu Lo Gwen Ray October 29, 2003 I. What is Natural Language Database Interface? • • allows users to access information stored in a database by formulating request in Natural Language. Examples: Who is the youngest employee in the sales department? Does any employee in the sales department earn more than $35000? Two Components • Linguistic component translates Natural Language input into a formal query and generating a natural language response based on the results from the database search. • Database component Performs traditional database management functions Simple Diagram of NLDBIS Natural Language Input Linguistic Component •Natural Language Understanding (morphological analysis, syntactic analysis, semantic analysis, discourse analysis) •Natural Language Generation Transfer to Database Language Database Component Response Linguistic component Includes : • Natural Language Understanding Mapping the given input in the natural language into a useful representation. • Natural Language Generation Producing output in the natural language from some internal representation. II. What is it related to, or what will affect? • companies are now making their product databases available online with a multitude of user interfaces. • their accessibility to no-expert users is desirable. III. Why use Natural Language Interface to Database? • Widespread use of Database • Accurate retrieval IV. What and how did it start? • 1973 -- The first NLDBIS was the LUNAR system built by Woods. It used chemical analyses of moon rocks. • By late 1970's -- first good database natural language processing systems was developed by LIFER/LADDER about US Navy Ships. This system used a semantic grammar to parse questions and query a distributed database. IV. What and how did it start? (cont.) • 1980-1990 -- NLDB systems became large, diverse and represented completely different concept. • Recent -- commercial products have created NLDB system to transfer written or spoken language into database queries e.g., PDAs, cell phones. Besides, multilingual natural language interfaces have developed in the e-commerce environments. V. Why is it important, and to whom? Database Users – want ease; reliability & accuracy of responses; reduction in time spent querying; global accessibility. Org.s/Companies – improved customer service, greater and global accessibility; cost to employ vs. savings. Database Developers & Administrators - Additional database development human resources for linguistic components development (including multiple languages); Additional database development testing for reliability and performance and security. VI. What is the current state? Web access Global access/multilingual Small devices – PDAs, Cell Phones Less specialized, more varied databases Non-specialized database users Multiple databases VI. What is the current state? (cont.) New natural language database interfaces for new users! New techniques/processes that will support the wide variety of databases being accessed by a wide variety of users, around the globe! An example: The Precise System designed by computer scientists at the University of Washington. VI. Current state The Precise System VI. Current state The Precise System (contd.) VI. Current state Other initiatives Spoken language dialogue systems. Tutorial systems. Large-scale dependency grammars which can include semantic and morphologic information, hierarchical statistical language models. Inductive logic programming to construct natural language processing systems (CHILL system – Mooney) VII. What do you expect in the immediate future? More transportable natural language interfaces to a wider spectrum of databases. Improved natural language interface used with improved voice recognition. More appliances with natural language interfaces. VIII. What are the implications for database users, administrators, developers, and the general public? Now that transportable NLIs are in development that make the use of them more feasible across a wider spectrum of databases, it is likely we will see an expansion of the use of natural language database interfaces. The trend is already afoot and if users expect it, it will likely come. VIII. Implications (cont.) Database Users Better or expanded customer service Ease of finding answers to questions Administrators Additional database development testing for reliability and performance and security Developers Additional database development human resources for linguistic components development (including multiple languages) General Public Improved Global accessibility The End. Questions?