Randomized algorithm for Information Retrieval using past search results

In Information Retrieval, past searches are a source of useful information for new searches. This paper presents an approach for reusing past queries submitted to an information retrieval system and their returned results to build the result list for a new submitted query. This approach is based on a Monte Carlo algorithm to select past search results to answer the new query. The proposed algorithm is easy to implement and does not require learning. First experiments were carried out to evaluate the proposed algorithm. These experiments used a simulated dataset (i.e., document collections, queries and judgments of users are simulated). The proposed approach was compared with a traditional approach of information retrieval, showing better precision for our proposed approach.