Dialogue system construction method and semantic controlled response method and device

Abstract

The invention provides a dialogue system construction method and a semantic controlled response method and device. A dialogue system comprises a first question-answer model and a second question-answer model. When a user inputs a question, firstly, an introducer associated with question semanteme is obtained from the question; then, the first question-answer model is used for generating a first response to the question, wherein the first response comprises partial response content of the dialogue system to the question; then, the second question-answer model is used for generating a second response to the question, wherein the second response comprises all responses of the dialogue system to the question. The introducer can control the output of a first response model and the output of a second response model, so that the semanteme of the first response and the semanteme of the second response are controlled. Therefore, by means of the semantic controlled response method, the semanticassociation degree of the responses generated for unknown questions and the questions is high; in addition, along with the change of the introducer, the responses generated for the unknown questions can also be changed, so that the diversity of the question responses is improved.

Description

technical field [0001] The present application relates to the technical field of natural language processing, and in particular, to a method for constructing a dialogue system, a method and device for semantically controlled response. Background technique [0002] With the development of natural language processing technology, dialogue systems established based on natural language processing technology have also been widely used. Commonly used dialogue systems such as chat robots can automatically generate corresponding responses according to the chat content input by users. [0003] Dialogue systems can be divided into target-driven dialogue systems and non-target-driven dialogue systems according to different application purposes. Among them, the goal-driven dialogue system refers to a dialogue system with clear service goals or service objects, such as customer service robots, weather broadcasting robots, etc.; non-object-driven dialogue systems refer to those developed w...

AI Technical Summary

Problems solved by technology

[0006] Although the dialogue system based on the deep learning model of the prior art can give an answer to an unknown dialogue, since the deep learning model of the prior art uses a probability-based algorithm in the field of statistics to realize the training process , after it is trained, what kind of response to an unknown dialogue is a probabilistic event with uncontrolled semantics. Therefore, the dialogue system based on the deep learning model in the prior art often gives different dialogues to the user. If the same general answer is given, the semantics of the answer cannot be controlled according to the user's dialogue content, and answers with different semantics are given, which leads to insufficient correlation between the answer and the question, and affects the user experience.

Images