Augmenting Data Retrieval with Information Retrieval Techniques by Using Word Similarity

Abstract

Data retrieval (DR) and information retrieval (IR) have traditionally occupied two distinct niches in the world of information systems. DR systems effectively store and query structured data, but lack the flexibility of IR, i.e., the ability to retrieve results which only partially match a given query. IR, on the other hand, is quite useful for retrieving partial matches, but lacks the completed query specification on semantically unambiguous data of DR systems. Due to these drawbacks, we propose an approach to combine the two systems using pre-defined word similarities to determine the correlation between a keyword query (commonly used in IR) and data records stored in the inner framework of a standard RDBMS. Our integrated approach is flexible, context-free, and can be used on a wide variety of RDBs. Experimental results show that RDBMSs using our word-similarity matching approach achieve high mean average precision in retrieving relevant answers, besides exact matches, to a keyword query, which is a significant enhancement of query processing in RDBMSs.