Using Clustering and Blade Clusters in the TeraByte task

[摘要] Web search engines exploit conjunctive queries and special ranking criteria whichdiffer from the disjunctive queries typically used for adhoc retrieval. We wanted toasses the effectiveness of those techniques in the TeraByte task, in particular scoringcriteria like: link popularity, proximity boosting, home page score, descriptions andanchor text. Since conjunctive queries sometimes produce low recall, we tested a newapproach to query expansion, which extracts additional query terms from a clusteringof the snippets from the first query. The technique proved effective, almost doublingthe Mean Average Precision. However, the improvement was just enough tocompensate for the drop that was introduced, contrary to our expectations, by the

[学科分类] 社会科学、人文和艺术（综合）