Processamento de consultas documento-a-documento utilizando índice em camadas

Abstract

Search engines are mechanisms to seek relevant information within large data collections. The constant growth of electronic media for storage information, along with the popularization of search engines, brings the constant need for solutions that reduce processing costs queries. We present two new algorithms for query processing searching systems. The processing algorithms use the approach document-to-document and modify the current algorithm state of the art, BMW, for taking advantage of an index architecture divided into two layers. The first layer contains only the highest impact index entries and is used to preprocess consultations before accessing the rest of the index in the second layer. This approach results in significant performance gains. The first algorithm, called BMW-CS, is up to 40 times more fast compared to many compared methods, but causes small changes in the returned response. The second algorithm, called BMW-t, preserves the answer set and is 10% faster than the BMW.