Automatic Quality Evaluation for User Generated Contents in Online Q&A Community Based on Word2Vec-CNN

The online Q&A (question-and-answer) community has emerged as a prominent platform for Internet users to access valuable information, but a notable disparity exists in the quality of answers provided. Hence, understanding and enhancing answer quality evaluation in the context of Chinese Q&A communities assumes paramount significance. First, a system of evaluation metrics for answer quality is constructed. Then, based on this metric system, the collected Zhihu data is annotated and an answer quality automated evaluation model using Word2VecCNN (convolutional neural networks) is developed. Utilize this model to perform both coarse-grained and fine-grained automated evaluation of answer quality in online Q&A communities. Compared to other models, the proposed Word2Vec-CNN-based automatic answer quality evaluation model in this research yielded superior experimental results, particularly in the realms of syntax, semantics, and pragmatics. Furthermore, it closely approximated the manually curated content found in “Zhihu Daily.” It excelled in filtering high-quality answers, significantly enhancing users’ efficiency in acquiring knowledge during their actual browsing experiences within the Q&A community. Consequently, it demonstrates a stronger potential to leverage the pivotal role of Q&A communities in disseminating network knowledge.