Answer Quality Prediction Joint Textual and Non-Textual Features

Community question answering (CQA) is a popular online service for people to ask and answer questions. But along with the increasing of user generating contents, the quality of answers provided by different users varies widely. So the quality of the answer caused wide attention. In this paper, we propose an answer quality prediction model to evaluate the answer quality considering both aspects of textual and non-textual features. We firstly employ Bidirectional long Short-Term Memory (BLSTM) based RNN model to evaluate textual quality of the answers. And we extract 11 features of the answers to evaluate the non-textual quality of answers. Finally, we jointly consider the score of answers textual and non-textual qualities. We evaluate our model in a benchmark dataset and the experimental results show that our model outperforms other existing approaches.