基于语义情感相似度的问答社区答案排序研究

Research on Ranking Q& A Community Answers Based on Semantic Emotional and Similarity

【目的/意义】目前问答社区答案数量多且篇幅长,对答案进行重新排序和组织可以方便用户高效获取答案信息。【方法/过程】针对答案文本的特点,提出基于答案相似度对社区答案进行重新组织、排序的方法。该方法定义答案文本语义云和情感云,利用标签传播算法计算出词典中不存在的词语语义和情感相似度,即云滴值,然后每个答案形成多片语义云和一片情感云,通过计算答案云间相似度得到答案文本间的相似度。最后,结合答案"赞"数对答案进行重新排序。【结果/结论】通过实验的人工评价,发现与基于"赞"数排序相比,基于语义情感相似度的答案排序方法与人工排序相似度更高,更能满足社区用户需求。

【Purpose/significance】There are a large number of-answers with lots of words in the QA community now. Thereordering and organizing of answers can facilitate the users to obtain theanswer information efficiently.【Method/process】According to the characteristics of the online QA community answers, this paper proposes the reordering of answers basedon similarity computing. This method defines the answer semantic cloud and emotional cloud,and uses the labelpropagation algorithm to compute the value of every cloud droplet, then combines two clouds into one, called answer cloud.The similarity of the text is obtained by calculating the cloud similarity. Finally, the answers are reordered according to thenumber of answers ＂like＂.【Result/conclusion】Through the artificial evaluation of the experiment, it is found that theranking method based on the similarity of semantics and emotion is more similar to manualranking than the ranking basedon the number of ＂like＂, which can better meet the needs of community users.