Here is a 2004 paper on Association Thesaurus Construction for Interactive Query Expansion based on Association Rule Mining

The article discusses basic association rule mining concepts like support, confidence, and pruning as we described in Association Rules Part 1 (July issue of IR Watch – The Newsletter). BTW, read Part 2 in the August issue.

The mentioned paper states:

“This paper presents an interactive query expansion method with association thesaurus, which is mined from the ‘selected web pages’ of users in the query logs. The ‘selected web pages’ of users are transferred into ‘sets of query terms’ and then used for term correlation mining. Accordingly, various association thesauruses concerning different query terms are constructed from these term correlations. Consequently, the proposed method combines the original query term specified by a user with the corresponding thesaurus to offer the user more precise results. The query expansion mechanism is implemented within the Query Agent of a course recommendation system, Coursebot. Experimental results have shown that the performance, precision ratio and recall ratio, of the system is increased when the proposed method is applied.”

The idea of mining query logs to construct association rules is not new, however their procedure is quite interesting.