Distributed Top-k Keyword Search over Very Large Databases with MapReduce

Abstract

In the last decade, keyword search over relational databases has been extensively studied because it promises to allow users lacking knowledge of structured query languages or unaware of the database schema to query the database in an intuitive way. The existing works about keyword search on databases proposed many approaches and have gain remarkable results. However, most of these approaches are designed for the centralized setting where keyword search is processed by only a single server. In reality, the scale of databases increases sharply and centralized methods hardly can handle keyword queries over these large databases. Moreover, processing keyword search over relational databases is a very time-consuming task, and the efficiency of the existing centralized approaches will degrade notably because the single server cannot provide enough computation power for the keyword search over very large databases. To address these challenges, we propose a distributed keyword search (DKS) approach with MapReduce and this approach can be well deployed on a cluster of servers to deal with keyword search over large databases in a parallel way.