Abstract:Q&A communities are playing an important role in online knowledge sharing, where a large number of users with various knowledge background make tremendous contributions to solving many technical problems based on crowd intelligence. However, as new questions are increasingly posted, it is a non-trivial issue to find a matching answerer for each question. As a result, many questions fail to receive satisfying answers in time. This paper addresses the problem by predicting the best answerer for the new question. Many existing efforts are devoted to predicting the best answerer mainly by calculating the textual similarity between questions and a user’s historical post documents. Some works consider other features, such as the similarity of tags between questions and users, the average quality of a user’s historical answers, and so on. But few works consider interaction within the community. In recent years, works that take account of the interaction between community items (such as GCN and GAT) have made considerable progress in graph mining tasks like item recommendation, node representation, node classification, and link prediction. This kind of graph mining method can easily leverage interactive information in the community and encode it in an easy-to-use way which is very helpful for downstream tasks such as recommendation. However, questions that need to be recommended to answerers are new coming ones and with no interaction with any other node in the community yet. How to make reasonable use of collaborative information to improve recommendation performance is a real challenge. In this paper, we use the interactive information between candidate answerers and combine text information to make our best answerer recommendation. There are two main parts, LDA(Latent Dirichlet Allocation) topic model is used to capture the text information and graph attention networks (GATs) for interaction. We evaluated our approach on a real dataset from Stack Exchange. The result shows that our approach outperforms all the baseline methods.

Answer Quality Characteristics and Prediction on an Academic Q&A Site: A Case Study on ResearchGate

Characterizing Peer-Judged Answer Quality on Academic Q&A Sites: A Cross-Disciplinary Case Study on ResearchGate.

Evaluating Academic Answer Quality: A Pilot Study On Researchgate Q&A

Researchers' judgment criteria of high-quality answers on academic social Q&A platforms

Quality Evaluation and Prediction for Question and Answer in Chinese Community Question Answering

Factors Influencing the Importance of Criteria for Judging Answer Quality on Academic Social Q&A Platforms.

Toward Personalized Activity Level Prediction in Community Question Answering Websites

Research on the Quality Prediction of Online Chinese Question Answering Community Answers Based on Comments.

Questions or Questioners: Factors Affecting Response Quantity on Academic Social Q&A Sites.

Exploring Social Features for Answer Quality Prediction in CQA Portals

Answer Quality Analysis on Community Question Answering

A Judgment Model for Academic Content Quality on Social Media.

Modeling information-seeker satisfaction in community question answering

An empirical assessment of best-answer prediction models in technical Q&A sites

Want a Good Answer? Ask a Good Question First!

Analysis of the Influencing Factors of Users’ Adoption Behavior in Social Q&A Community Based on Machine Learning Regression Algorithms

The Dynamics of Q&A in Academic Social Networking Sites: Insights from Participants, Interaction Network, Response Time, and Discipline Differences

A deep learning approach for predicting the quality of online health expert question-answering services

Answer Evaluation In Q&A Community Site On The Topic Of Natural Disaster

Best Answerers Prediction with Topic Based GAT in Q&A Sites.

Predicting information seeker satisfaction in community question answering