Ranking Tagged Resources Using Social Semantic Relevance

1. Introduction

The World Wide Web has evolved as a unanimous powerful information portal that allows a learner to retrieve and access available resources or documents on any subject or topic. Various search engines, with their rational design and organizational efficiency retrieve web pages on a user’s query or search topic. However, the resulted web pages are usually very large in number which consequently leaves the user with the only option to search required content by visiting each web page. Such large volume of search results therefore needs a ranking according to the user’s information requirements. The ranking algorithms used to rank search results often relay on the formation of user’s query, and its Boolean and syntactic keyword matching in the web pages’ content. This means that the document / resource or web page relevance is determined by using the words in the topic or the query which is matched (keyword based) with the content of the web pages. The web page retrieval and ranking algorithms (Yi, Liangjie, Ruihua, Jian-Yun, & Ji-Rong, 2009; Chawla & Bedi, 2008; Greg, Chowdhury, & Torgeson) based on classification, clustering of queries and documents, click logs, terms co-occurrence (Matsuo and Ishizuka 2003), etc., are the few used by many search engines. Apart from these, the algorithms based on the term frequency (TF, IDF) (Salton and Buckley 1987), internal, external links, and clickthrough of HIT count methods (Kleinberg, 1999; Kazunari, Kenji, Masatoshi, & Shunsuke, 2003) with their various modified versions, that uses syntactic keyword matching (Agrahri, Anand, & Riedl, 2008) are also used to search and rank relevant documents. These algorithms work best for a well formed query with specific keywords because of their syntactic searching potential. However, the ranking criterion built on the algorithms like clickthrough usually gets biased interpretations in selecting links. This is apparent from the study (Chi and Mytkowicz 2008) that shows more than 90% of search sessions do not go beyond the first result page. As the result, most of the time, the users click links available only on first few pages, ignoring various links that exist at the end pages. Moreover, people have also been found biased in the selection of search engines and their search activities (Shenghua, et al. 2007). Besides this, many links that exist at end pages never get a chance to get their HIT. All above studies infer that the users’ clicks are biased in nature and are influenced by many factors like placing few links at the top on the first page of search results, selection of search engine and placing popular keywords in the web pages. However while computing relevance to rank resources, if semantically related terms to the search topic or query and some sort of the community’s feedback regarding web resources are incorporated, the ranking order of resources can be improved.

This paper proposes Social Semantic Ranking (SSR) to rank tagged web resource using Social Semantic Relevance (S2R) which is based on Vector Space Model. It uses tags which are bookmarked in Social Bookmarking Site (SBS), and pre-existing semantic knowledge from ontologies for computing Semantic Relevance of the crawled web resources on a search topic.

The rest of the paper is organized as follows: In Section 2 the related work and some basic features used in the approach have been discussed. The proposed approach for Social Semantic Relevance ranking of tagged resources with respect to a search topic is explained in Section 3. Section 4 presents the implementation of the approach and evaluation of its ranked results in the experimental study. Finally Section 5 concludes the paper.