An answerer recommender system exploiting collaboration in CQA services

Abstract

Community-based Question Answering (CQA) services are becoming popular as the public gets used to look for help and obtain information. Existing CQA services try to recommend someone for answering new questions. On the other hand, people are allowed to exchange information and experience using various collaborative tools. It would be interesting to combine the two approaches to increase the reliability of recommending an answerer. Thus, relying on semantically modeled traces, we propose a comprehensive approach that recommends an answerer in a collaborative environment. From a global point of view, this approach consists in evaluating users by the performance in the CQA services and the corresponding knowledge sharing activities in which they participated in a collaborative context. By modeling and analyzing users' behavior, we assess the competency of an answerer in a particular collaborative context.

A Survey on Expert Recommendation in Community Question Answering

“…However, they leverage users' voting information on answers as the "relevance" labels and utilize LambdaMART to learn ranking models which directly optimizes a rank-based evaluation metric, normalized discounted cumulative gain (nDCG). Logistic regression [76] has also been used recently to facilitate both ranking and classification.…”