摘要
问答社区中问题所带有的标签能够帮助用户高效地查找和共享信息资源,而问题文本相较于一般文本所含的统计信息更少。现有的标签推荐算法大多基于统计方法,难以很好地表达问题文本所包含的特征,进而导致标签推荐效果较差。针对上述问题,提出一种基于共现图协同过滤的混合标签推荐算法。该推荐算法主要由共现图构建模块、词向量模型协同过滤模块以及投票融合模块组成。在知乎问答社区互联网话题的问答数据集上的实验结果表明,提出的混合标签推荐算法具有更好的推荐效果。
Tags of question in Community Question Answering(CQA)can help users find and share information,but the question contents contain less statistical information than general documents.Since most recommendation models are based on statistical information,the effectiveness of these models used in tag recommendation is not very high.To address this issue,a hybrid tag recommendation algorithm based on co-occurrence graph is proposed.This algorithm consists of co-occurrence graph construction module,collaborative filtering module based on Word2 Vec model and voting fusion module to generate final tag recommendation list.By conducting experiments on the question-answer(QA)dataset under topics of Internet in Zhihu,the results show that the recommendation algorithm has higher effectiveness.
作者
田伟
龚磊
TIAN Wei;GONG Lei(College of Computer Science,Sichuan University,Chengdu 610065)
出处
《现代计算机》
2020年第19期40-44,共5页
Modern Computer
关键词
问答社区
标签推荐
共现关系
词向量模型
Community Question Answering
Tag Recommendation
Co-Occurrence
Word2Vec
