Entities, Terms and Phrases for Recovering Dead Broken Links (T-0735) (MFN 6879)

Abstract

Web is of dynamic nature, lots of changes occurred daily. Increase in new Web sites and updating, deleting, or modification make Web more dynamic. The rapid changes in Webpages increase the broken links, the user face the broken links when they search of required information. Several approaches are presented by research community to provide solution to the problem. The approaches, discussed in chapter 2, used various techniques for recovery of those broken links that includes, source page, anchor text, URL and surrounding text around anchor text. In the thesis, the research study is based on data mining approaches, to identify the most suitable query terms for broken links. In Pervious researches not all the words/terms in the pages text can useful and source page text can efficient enough to retrieve the broken page. During the research work it is ana lysed that not only single terms but phrases and entities in the page text have increase the chance to recover the broken links. Therefore, to prove the approach initially the terms, phrases and entities are extracted from the source pages and classify them as good and bad query terms. Manually these queries are assigned to search engine "Google". Furthermore, data mining approaches are applied to classify the extracted query candidate into two classes (i.e. good and bad). The results reveal that among the three sources of queries i.e. terms, phrases and entities; only phrases proved good source for query as compared to other.