Broadening Vector Space Schemes for Improving the Quality of Information Retrieval

摘要

The vector space model (VSM) of information retrieval suffers in two areas, it does not utilise term positions and it treats every term as being independent. We examine two information retrieval methods based on the simple vector space model. The first uses the query term position flow within the documents to calculate the document score, the second includes related terms in the query by making use of term correlations. Both of these methods show significant improvements over the VSM precision while keeping the query time to speeds similar to those of the VSM.