Achieving Effective Multi-term Queries for Fast DHT Information Retrieval

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Distributed Hash Tables (DHTs) are well-suited for exact match look- ups using unique identifiers, but do not directly support multi-term queries. Related research of query expansion has shown that adding new terms to a query via ad hoc feedback improves the retrieval effectiveness of such query. In the paper, we propose an effective multi-term query processing algorithm for information retrieval in DHT systems. Given the significance of first term in a multi-term query, the query is sent to the peers containing the first term. To enhance the query effectiveness, we design two query expansion mechanisms and an implicit relevance feedback approach based on users’ behaviors. Additionally, we record the query log and the expansion terms for each query which can accelerate the future queries and improve the query accuracy. Experimental results show that our query methods yield substantial improvements in retrieval effectiveness in the following three aspects: recall, precision at 10 standard recall levels and precision histograms.