Algorithms for community subject-identification: experiments and test results

Abstract

We deal with the problem of finding the subject that a given set of pages deals with. The algorithms presented have been designed with a view to provide good results when pages that belong to community (or handle some common topic) are presented to them. When a set of pages are presented, the algorithms return a set of words ordered in the decreasing order of relevance. The words among the first few can be used to identify the topic that the set deals with. The algorithms have been tested by providing them with pages returned from a search engine for a given search query and their performance is assessed based on how high the words used in the search query are placed in the list returned by the algorithm.