摘要
Abstract
Under the background of the rapid development of the Internet,community Q&A website questioners have a stronger demand for knowledge.Massive data poses difficulties for questioners in identifying effective information,making it particularly important to recommend more professional expert users to answer questions.However,it is difficult to accurately calculate the correlation between the object question raised by the questioner and the candidate experts by the traditional community Q&A expert recommendation methods.In order to improve the efficiency of expert recommendation in community Q&A websites,an undirected graph of the problem node relationship is constructed,and a graph neural network(GNN)GraphSAGE is used to extract the second-order neighbor information of nodes.The multi-view learning method is used to learn the complementary information between different views and finally obtain a rich vector representation of the object question text and the candidate expert's historical question set,which is used to calculate the matching degree between the object question and the candidate expert,and then recommend the most suitable expert user to answer the object question.The experimental results show that,in comparison with different community Q&A expert recommendation methods,the proposed method has achieved better recommendation results on both the evaluation indicators MRR and NDCG@10.关键词
社区问答/专家推荐/图神经网络/多视图学习/推荐系统/深度学习模型Key words
community Q&A/expert recommendation/GNN/multi-view learning/recommendation system/deep learning model分类
信息技术与安全科学
