Semi-Supervised Evaluation of Search Engines via Semantic Mapping (2003)
| Citations: | 3 - 0 self |
BibTeX
@MISC{Menczer03semi-supervisedevaluation,
author = {Filippo Menczer},
title = {Semi-Supervised Evaluation of Search Engines via Semantic Mapping},
year = {2003}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Content and link information is used by virtually all search engines to crawl, index, retrieve, and rank Web pages. The correlations between similarity measures based on these cues and on semantic associations between pages is crucial in determining the performance of any search tool. A great deal of research is under way to understand how to effectively extract semantic information from Web pages by mining their text and links. A brute force approach has been used to build semantic maps, that given coordinates based on text and link similarity measures between two pages estimate how the two pages are related in meaning. In this paper we describe a semi-supervised methodology to evaluate the performance of search engines. The relevance of entire hit lists can be estimated by identifying a single relevant page for a given query. We illustrate the evaluation methodology by graphing precision-recall plots for three commercial search engines based on TREC queries. Preliminary results are in line with our intuition of how the different search engines use content and link information to rank hits.
Keyphrases
search engine semi-supervised evaluation semantic mapping link information entire hit list precision-recall plot search tool commercial search engine different search engine semi-supervised methodology great deal semantic map evaluation methodology preliminary result similarity measure single relevant page brute force approach trec query link similarity measure semantic association rank web page semantic information web page
