Abstract

Searching for documents using search engines is done by comparing the query submitted by the user to the documents in the collection. The submitted query contains one or more words carefully selected to retrieve as many relevant documents as possible. There have been many approaches that performed query expansion using synonyms. Retrieval effectiveness can be increased using query expansion, like the query term "PC" can be expanded with "Laptop", "Desktop", and "Computer", then documents containing these synonyms will also be retrieved. This paper aims in introducing a new model that mines association rules from the document collection itself using the Apriori algorithm and expands the query with words learned from association rule mining. Confidence values for association rules are used, where terms with high Confidence will be added to our query, which are learned from association rule mining. The query without any expansion will be our baseline query. We use a vector space model with TF‑IDF weighting and cosine similarity. The proposed technique is evaluated on the Medline test collection, consisting of 1,033 documents and 30 queries. Queries are first run as-is, then expanded using associated terms, and the results are compared using Precision and Recall. This approach can complement existing query expansion methods by adding terms automatically discovered from the document collection itself, without relying on external knowledge sources.

Statistics