Finding Instantaneous Community of Ideologically Similar Users in Social Forums

The concept of community in online Social Forums is important and many times content shared by individual disseminated across the group members or friends based on the individual preferences and social forums policies. The typical friend suggestion algorithms used by many Social Forums depend on acquaintance with others users, e.g. because of relationships such as family, school mates, and professional etc. In this paper, we have proposed an algorithm for friend suggestion based on ideological affinity among the users, who sometimes even might not have direct communication amongst each other. The proposed algorithms makes use of Cosine Similarity for deriving similarity score, for a intersection of topics greater than or equal to a threshold factor, where, threshold factor is a natural number which represents minimal no. of elements in intersection of sentiment vectors (a map containing topics and its sentiment scores) of two users, such that similarity algorithm can be applied for calculation of similarity score. We have derived sentiment score in an unstructured message using Stanford Core NLP API and we have further performed clustering of users who have semantic linkage/ ideological similarity/ideological affinity. The created cluster represents instantaneous community or live community at the time of ongoing discussion based on the data fetched in previous few seconds; such live community members can connect with each other and can discuss on topics on similar interests and enrich their experience, further topics can evolve together with given social interactions between participants and they can also work for solving problems on some common cause.