Generation method of professor intention answers in man-machine conversations

Abstract

The invention discloses a generation method of professor intention answers in man-machine conversations. The method comprises the steps that according to received conversations, sentence similarities are calculated, and then dialogue intention recognition is performed on the current dialogue sentences; if the current dialogue intention is a small talk, the current dialogue and a corresponding inquiry intention value corresponding to the current dialogue are added into historical records, and answers are directly returned by searching a knowledge base or a network; if the current dialogue intention is a professor, the next step is executed; the historical records are searched for questions corresponding to the current dialogue; the current dialogue and network information are combined, and a plurality of simulation internal dialogues are performed to obtain a related answer set; the related answer set is filtered; on the basis of each answer weight, an abstract is extracted, the answer with the highest weight is adopted as an abstract extraction result and the current dialogue answer to be returned. According to the method, the professor content of a user can be well fed back under the professor intention in man-machine conversations, and the satisfaction of the man-machine conversations is improved.

Description

technical field [0001] The invention relates to the technical field of man-machine dialogue, in particular to a method for generating answers to teaching intentions in man-machine dialogue. Background technique [0002] Human-computer dialogue is an important research topic in the field of artificial intelligence. During the dialogue process, user experience is very important. With the development of speech recognition technology, human-computer dialogue plays an important role in various fields such as chat robots, voice assistants, virtual human scene dialogues, and children's chat robots. However, in the process of man-machine dialogue, for some questions, the answer of the robot is not correct, or the user holds different opinions on the answer returned by the robot. At this time, the user often expresses his opinion to correct or teach the robot. In this case, what the user hopes more is that the robot can answer his point of view, but the current man-machine dialogue ...

AI Technical Summary

Problems solved by technology

However, in the process of man-machine dialogue, for some questions, the answer of the robot is not correct, or the user holds different opinions on the answer returned by the robot. At this time, the user often expresses his opinion to correct or teach the robot. In this case, what the user hopes more is that the robot can answer his point of view, but the current man-machine dialogue system often returns the answer to the knowledge base directly or gives an ambiguous reply, although such a reply can reduce the computer's positive feedback. The answer and the error rate of the answer, but it gives the user the feeling that the computer did not understand the user's communication or the intention of the professor. Such a dialogue experience is not friendly

Images