Aiming at the questions not answered timely under Q&A community, a kind of questions recommendation method based on LDA (Latent Dirichlet Allocation) topic model is proposed, which fully utilizes personalized information of users under Q&A community. The interests distributions of users are expressed through using LDA model and according to the interests distributions of users, questions recommendation lists are calculated out at last. The proposed method can recommend the unsolved problems to users who are interested in these questions, which makes these questions be solved out as soon as possible, and promotes information dissemination and knowledge sharing under Q&A community. Experimental results show that the proposed questions recommendation method based on LDA not only discoveries the unsolved questions quickly, but also recommends the most suitable answers to users compared with PLSA, KL-divergence and Cosine similarity.

Because the accuracy of traditional sentiment orientation identification algorithm is not high under Q&A community, this paper proposes a new method based on two-level conditional random field improved by particle swarm optimization algorithm for emotion tendency recognition under Q&A community. The proposed method adopts particle swarm optimization algorithm to train two-level conditional random field model, and applies the trained conditional random field model to recognize emotion orientation of question-answer pairs in Q&A community. Experiments were performed on Yahoo! Answers data set and results show that the proposed two-level conditions random field improved by particle swarm optimization algorithm has a higher precision rate, recall rate and F1 value at the micro average and macro average aspects compared with Hidden Markov Model, Max-Entropy Markov Model, Support Vector Machine and traditional condition random domain model, which prove the proposed two-level conditions random field improved by particle swarm optimization algorithm is a more effective method to recognize emotion orientation of question-answer pairs in Q&A community.