The notable feature of a question answering system is to understand the semantic information of the user’s question. Question classification, as the key module of question answering system, plays a decisive role in the efficiency of system retrieval. In order to classify the user’s questions, a classification model of tomato pests and diseases based on word2vec and bidirectional gated recurrent unit (BIGRU) was constructed. word2vec was used to transform the words in the sentence into the word vector with semantic information. The word vector was used as the initial corpus. Two neural network methods and a machine learning method were adopted to train the classification model. Totally 2000 tomato pests and diseases related questions were selected, which were divided into two categories: tomato diseases and tomato pests. The results showed that the classification accuracy, recall rate and F1 value by using the BIGRU model were 2~5 percentage points higher than those by using convolutional ceural network (CNN) and K-nearest neighbor (KNN) classification algorithm. Further experimental results comparison indicated that the BIGRU model performed the best on tomato pest and diseases question classification. The BIGRU model was simple in structure, less in model training parameters, and fast in training speed. It met the response time requirements of question answering system.