Query Planning for Searching Inter-dependent Deep-Web Databases

  • Authors:

  • Fan Wang;Gagan Agrawal;Ruoming Jin

  • Affiliations:

  • Department of Computer Science and Engineering, Ohio State University, Columbus OH 43210;Department of Computer Science and Engineering, Ohio State University, Columbus OH 43210;Department of Computer Science, Kent State University, Kent OH 44242

  • Venue:
  • SSDBM '08 Proceedings of the 20th international conference on Scientific and Statistical Database Management
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Increasingly, many data sources appear as online databases, hidden behind query forms, thus forming what is referred to as the deep web. It is desirable to have systems that can provide a high-level and simple interface for users to query such data sources, and can automate data retrieval from the deep web. However, such systems need to address the following challenges. First, in most cases, no single database can provide all desired data, and therefore, multiple different databases need to be queried for a given user query. Second, due to the dependencies present between the deep-web databases, certain databases must be queried before others. Third, some database may not be available at certain times because of network or hardware problems, and therefore, the query planning should be capable of dealing with unavailable databases and generating alternative plans when the optimal one is not feasible.This paper considers query planning in the context of a deep-web integration system. We have developed a dynamic query planner to generate an efficient query order based on the database dependencies. Our query planner is able to select the topKquery plans. We also develop cost models suitable for query planning for deep web mining. Our implementation and evaluation has been made in the context of a bioinformatics system, SNPMiner. We have compared our algorithm with a naive algorithm and the optimal algorithm. We show that for the 30 queries we used, our algorithm outperformed the naive algorithm and obtained very similar results as the optimal algorithm. Our experiments also show the scalability of our system with respect to the number of data sources involved and the number of query terms.