Typer av sökfrågor på webben
En effektivitetsstudie
The purpose of this thesis is to examine the performance of different types of queries in the AltaVista query-based search engine. Thirty information needs have been collected from an ask-a-question service on the World Wide Web, which means that they are based on real users information needs. This is done to avoid partiality in the study. The author has invented different types of queries for each and every one of the information needs in accordance with the search syntax available for AltaVista. The types used are Boolean expressions, queries including proximity operators, phrases and queries directed towards a specific field. The ten first hits for each query type have been evaluated for relevance on a three-point scale. Irrelevant, inactive links or duplicates are given the value 0, partially relevant documents are given the value 0,5 and highly relevant documents are given the value 1. Special criteria are used as to prevent that these relevance judgements are done in a partial manner. The measure used is precision and to give credit to those query types that present relevant documents early in the ranked list, the average precision is calculated instead of traditional precision. The results show that queries including proximity operators are the most effective with respect to precision, followed by Boolean expressions, phrases and queries directed towards a specific field. The proximity queries are also assigned the highest score for relevancy, which means that this query type in this study is the best at retrieving relevant documents.