Community question answering (CQA) becomes increasingly prevalent in recent years, providing platforms for users with various background to obtain information and share knowledge. However, there are a large number of answers, difficult for users to view item by item and select the most relevant one. Therefore, answer selection and duplicate question detection become very significant subtask of CQA.
In this work, we propose different models to explore these tasks. First, we study the correlation between question and paired answer. Then, we introduce the attention-based model Himu-QAAN for the answer selection task. Also, we present a BERT-based model Bert-QAnet for duplicate question detection task. We test our methods on various datasets. The results show that our methods achieve significant performance.