Optimizing Ranking Calculation in Web Search Engines: a Case Study

Web search engines compete to offer the fastest responses with highest relevance. However, as Web collections grow, it becomes more difficult to achieve this purpose. As most users tend to see only the first two pages of results, it is unnecessary to compute the ranking of each one of the millions of documents that usually match any given query. Only those that have a level of importance that makes them candidates to the top ranked results have to be considered. This work presents and compares algorithms tested in our Web search engine to speed up the search of these candidates. We have been able to reduce by 93% the number of documents considered for ranking calculation, using a pruning algorithm over a Web collection index sorted by URL weights.

• Type: Conference
• Year: 2004
• Where: SBBD
• Authors: Miguel Costa, Mário J. Silva
• Keywords: SBBD 2004, SBBD 2007, Search Engines Compete, Web Collection, Web Search Engine