Multi-round Dialogue Intention Recognition Method for a Chatbot Baed on Deep Learning

With the continuous development of human-computer dialogue system, more and more dialogue robot products come into people’s lives. However, when human beings use short sentences and omit words, and in the process of identification often face problems such as more text noise, sparse characteristics, polysemy, backward and backward dialogue information. In order to solve the above problem, a deep learning based chatbot multi-round dialogue intention recognition method, according to the fit of deep learning algorithm and chatbot multi-round dialogue intention recognition model, by transforming the problem into a mathematical model, and obtain the final dialogue intention through the calculation of the model. First, the chat dialogue text was preprocessed, and the BERT model was established based on the processing results, the BERT model fused the deep learning model in the B E R T model, established a joint model, and data vectorized the short text of the human-computer dialogue. Finally, the multi-round dialogue intention identification similarity is calculated through the robot, realizing the dialogue intention recognition, and experiments show that the highest accuracy of the recognition method can reach 0.9912, the highest recall rate can reach 0.9914, and the highest f price is 0.9914, which can prove the superiority of the design method.