Supporting ontology-driven Keyword Search over Relational Databases

Abstract

Keyword Search Over Relational Databases (KSORDB) provides an easy way for casual users to access relational databases using a set of keywords. Although much research has been done and several prototypes have been de veloped recently, most of this research implements exact (also called syntactic or keyword) match. So, if there is a vocabulary mismatch, the user cannot get an answer although the database may contain relevant data. In this paper we propose a system that overcomes this issue. The proposed system extends exist ing schema-free KSORDB systems with semantic match fea tures. It exploits domain ontology to progressively return re lated terms that can be used to retrieve more relevant answers to user. Experimental results show that the semantic search method, employed by the proposed system, is more effective than the traditional keyword search method, employed by the existing schema-free KSORDB systems, in terms of the recall rate of the retrieved results.A signiticant amount of plain text and structured data has been stored side by side in relational databases for decades. In order to query this data, users have to know the database schema and then use Structure Query Language (SQL) to issue a precise, unambi guous and well-formed query. To relieve users from doing this, Keyword Search over Relational Database (KSORDB) enables casual users to query this data using a set of keywords called key word query. Existing KSORDB approaches can be categorized into two main categories: Schema-based approaches [1-3] and schema-free approach [4][5][6][7]. In order to process a keyword query, the schema-based approach uses the schema graph to generate Candidate Network (CNs) and then evaluate these networks using SQL queries. While in the second approach, schema-free, the data base is modeled as a data graph where nodes are tuples and edges are foreign -primary key relationships. The graph is then tra versed, at the query time, to answer keyword queries.The main difficulties with the schema-based KSORDB sys tems are in generating optimal SQL queries from a huge number of CNs. Moreover, the generated queries usually contain many join operations which are characterized by complex processing and time consumption features. Schema-free KSORDB systems have three main drawbacks: First, they are not efficient for large size databases as they consume the memory in materializing and processing the data graph. Second, the data graph should be up dated each time the underlying database is updated; therefore, this model is not appropriate for databases that change frequently. Third, traversing a huge data graph to tind minimum Steiner trees is proved to be an NP-hard problem [8]. Furthermore, both KSORDB approaches implement exact keyword matching/ search without exploiting the semantic relationships, such as such as meronym, hyponymy, and antonym, between keywords to find the correlated data.In Information Retrieval (IR) community, Ontology is used to support semantic matching. "An ontology is an explicit specifica tion of a conc...