Namesake alias mining on the Web and its role towards suspect tracking

Abstract

With the proliferation of social media, the number of active web-users is rapidly increasing these days. They create and maintain their personal web-profiles, and use them to interact with others in the cyber-space. Currently two major problems are being faced to automatically identify these web-users and correlate their web-profiles. First is the presence of namesakes on the Web, and the second is the use of alias names. In this paper, we propose a context-based text mining approach to discover alias names for all the namesakes sharing a common name on the Web, and leave the task of selecting the namesake of interest on part of the user. The proposed method employs a search-engine API to retrieve relevant webpages for a given name. The retrieved webpages are modeled into a graph, and a clustering algorithm is applied to disambiguate the webpages. Thereafter each obtained cluster standing for a namesake is mined for alias identification following a text pattern based statistical technique. The existing research works do not consider the presence of namesakes on the Web to mine aliases, which is impractical. The novelty of the proposed approach lies in discovering this drawback of existing works. Additionally the contribution includes the disambiguation technique that does not need to have a pre-determined number of clusters to be generated and the light-weight text pattern based alias mining technique. The number of clusters in the proposed method is rather determined dynamically by the inflation parameter, the pre-determination of which is comparatively much easier. Experimental results on different components demonstrate the robustness of the proposed alias mining approach. This paper also brings forth the significance of alias mining to the problem of suspect monitoring and tracking on the Web.