Personalization of Web Search Using Social Information

Abstract

In the recent years, there has been a rapid proliferation in the size of internet as the advent of web 2.0 made an end-user capable of generating various kinds of data by means of Interactive Internet Applications (IIAs), i.e., Facebook, Instagram, twitter, etc. This, in turn, posed a big challenge for the web search platforms to assist the web users in obtaining their desired information. The implication which further made the problem critical for search platforms is diversity in user's outlook towards the same thing. Therefore, to maintain the e ciency of search platforms, user's position must also be strengthened in web search through personalization in content generation. The work, presented in this thesis, proposes a new model for personalization of web search with a focus on the selection of composition corresponding to various supporting modules of an e cient personalization system. Collaborative tagging is one of the applications of IIA that facilitates a web user to annotate a web resource with a tag of interest. It is a rst-hand information directly given by a user without any middleman modi cation, therefore, it is more reliable than any other source. Thus, this collaborative tagging information can be quite helpful in constructing User Interest Pro le (UIP) and Resource Illustration Pro le (RIP). UIP will provide a complete list of user preferences along with his level of interest in that preference, while RIP enlists the topics about which a web resource describes or is related to, along with the degree of a nity for that topic. But the UIP constructed solely on the basis of user's own information is sparse and needs to be enriched with additional information. However, in the case of resource pro le, RIP constructed through collective information from all the users is ambiguous as every user holds a di erent viewpoint or feeling towards a web resource. The conventional methodologies have failed to redress these problems. The proposed model focuses on UIP enrichment using two di erent strategies. First one is clustering of tags based on the concept of semantic relatedness between two tags in the real-world. This has been measured using Word2vec model. The second one is the utilization of user's real society relationship network. It is believed that the present work is the rst one to integrate the concept of semantic relatedness for tag clustering. A novel approach has also been designed to handle outlier tags which caused ambiguity based on the concept of collaborative ltering. Even a good UIP and RIP alone cannot create an e cient personalization system, they also require a suitable mapping with user's query requirements. Therefore, in the proposed model, the fuzzy satisfaction requirementbased novel mapping functions have been designed to measure query relevance score and user interest relevance score for a web resource. These scores have been further used to calculate the post-relevance score of a web resource after a suitable trade-o . Unlike