Man-machine multi-round dialog model construction method based on scene contexts

Abstract

The invention provides a man-machine multi-round dialog model construction method based on scene contexts. The method comprises the operation steps that (a) abstract expression concepts of the three contexts of terms, question-answer sentences and scenes are defined; (b) based on extraction methods of the three contexts, an original question-answer base is trained to be a question-answer scene knowledge base represented by multiple scenes; and (c) for an input question, abstract representations of the three contexts of the question are acquired first, then scene matching is performed, a question-answer sentence context sequence in the scenes is extracted and combined with current question sentence contexts to form a new question-answer sentence context sequence, so that a reasonable answeris generated. Through the method, the importance of context information in the man-machine multi-round dialog process is fully considered, the answer can be successfully generated in combination withthe context information without causing information loss, the question-answer sequence conforming to the current scene can be extracted from complicated data to serve as guidance knowledge assistinganswer generation, and noise influences brought by useless data is greatly lowered.

Description

technical field [0001] The invention relates to the field of computer artificial intelligence natural language processing, in particular to a method for constructing a human-machine multi-round dialogue model based on scene context. Background technique [0002] The concept of intelligent chat robots has become more and more popular in recent years. Chat robots with voice interaction and natural language understanding as the core are gradually being applied to various service fields such as voice assistants, online consultation, guidance, customer service, and daily chat. [0003] However, most of the existing chat robots on the market have the common problem of often not answering questions. For a single round of question and answer, chatbots can obtain a more reasonable answer through retrieval, traditional semantic analysis and other methods; for long-term multi-round question and answer, chatbots often cannot grasp multiple topic information in the dialogue process, As ...

AI Technical Summary

Problems solved by technology

In the process of chatting, if the context content is completely discarded, the chat robot will not be able to correctly answer a large number of questions; and if all the context content is used, it will not only be complicated to deal with, but also introduce a lot of noise information, resulting in disordered responses. happened

Images