Efficient keyword proximity search using a frontier-reduce strategy based on d-distance graph index

  • Authors:

  • Ming Zhong;Mengchi Liu

  • Affiliations:

  • Wuhan University, Wuhan, China;Carleton University, Ottawa, Canada

  • Venue:
  • IDEAS '09 Proceedings of the 2009 International Database Engineering & Applications Symposium
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Current keyword proximity search approaches on general graph lack effective means to reduce the search space, and thus suffer from low efficiency when dealing with large search space. In this paper, we present a novel approach in order to address this problem. Our approach employs a best-effort frontier-reduce strategy that aims to find a set of subgraphs containing the best answers. So we need only to search over these small subgraphs to get the top-k answers, and thus the efficiency can be significantly improved. To fulfill our strategy, we define a d-distance subgraph with upper size bound, and extract such subgraphs from the graph to build a new index structure combining the mappings between keywords, vertexes and subgraphs, by which we can quickly look up the target subgraphs for specific queries. Then, we perform an efficient algorithm to find the top-k answers, which can overcome the subgraph overlap problem and support existing optimal prioritization techniques. Lastly, we evaluate the effectiveness and efficiency of our approach with extensive experiments. The experimental results show that our approach can outperform existing approaches by a large margin with little or none loss of answer quality.