Two-Stages Evaluation Algorithm for Automatic Ranking in Information Retrieval

摘要

First, it is described a prototype on an Information Retrieval system, GUMSe. The system sends user queries to multiple search engines and it gets a final ordering of documents. In this paper an information fusion technique is presented. GUMSe retrieves and merges the resulting URLs in an efficient way. The size of the document collection to be analysed is limited by a previous algorithm called "fast". Afterwards each document is retrieved, an exhaustive algorithm evaluates each document analyzing its content. This algorithm uses n-grams to assign a weight to terms semantically related with the original user query terms. The final score is based in the score of the two previous approaches.