Web retrieval: Techniques for the aggregation and selection of queries and answers

Reviewer: Fazli Can

In an information retrieval environment, we can cluster queries according to terms, operators, most relevant results, and other appropriate information. We can use information about old queries to improve the results for new queries. For better decision making, an authority may fuse the opinions provided by different experts by paying attention to their reliability. Within the context of this paper, the authority is an information retrieval system, the different experts are the clustering algorithms, and the opinions are the ordering (ranking) of the queries and their results in these clusters. The similarity of existing clusters to a new query can be interpreted as the reliability of clusters. This approach exploits the information provided by old queries to improve the retrieval performance for new queries. The authors propose a methodology that employs three different clustering algorithms and an aggregation method for data fusion in an environment as defined above. The methodology can be effective; however, it needs empirical evaluation, which is missing from the paper. As the next step to the work, the authors propose testing the approach using a query Web log. Online Computing Reviews Service