Webpage Duplicate Detection Using Combined POS and Sequence Alignment Algorithm

Abstract

Combined syntactical categories and sequence alignment algorithms are implemented and used to weed-out duplicate and near-duplicate web-pages from search engine results. The syntactical structures manifested as POS-tags were pre-processed using a POS tagger converting parts of a webpage's text into a string of tags. The produced string was then subjected into the longest Common Sequence (LCS) techniques (as is commonly done in computational biology), to detect duplicate and nearduplicate webpages. The process of tagging and aligning was based on set of sentences extracted from the web page as a representative of the pages. The query-keywords are used as a basis for sentence extraction. Results obtained from experiments performed have shown that such a combined approach can provide very interesting similarity calculation and re-ranking measure. This can be used with reasonable efficiency to detect duplications on search results generated by search engines such as Google. Similarity measurements obtained can be further used as a basis for text analysis of the search results allowing the detection of duplicate and near duplicates and clustering of documents in general.