EXPERT RECOMMENDER SYSTEM

Abstract

In online question & answer communities, People seek for expert opinions on their problems and share their expertise in a particular field. Hence expert finders are the building blocks of online Q&A communities and finding expert in such communities is one of the biggest challenges and an highly active research problem. Expert finding is the process of finding expertise level of each user and identifying erudite people on a given topic. Usually experts are find out in two ways - social network analysis or concept map. The recommendation system we have proposed is called Expert Recommendation system which incorporates two important features of recommender systems that are trust and similarity achieved by using a well known global trust metric Page rank and collaborative filtering respectively. For collaboration of experts, Firstly it is necessary to find experts in the community using social network analysis by using a variation of pagerank which will determine the expertise of a particular user by determine the expertise of users whom he has helped and score of the posts, he has contributed to. Spearman's rho is applied to these two lists of top-k users( one extracted from the community and other calculated by applying pagerank) to determine the correlation between them. Once, we have top-k experts of a community, using collaborative filtering we will recommend these experts to each other. Stack Overflow is a classic example of online Q&A Community which has 14M questions, 22M answers, 58M comments, 49K tags and 73M users till date. It assigns Reputation points to each of its users. So, We have used stack overflow to test our framework.