Session message processing method and device

Abstract

The invention provides a session message processing method and device. The method comprises: according to a target word in a session message inputted by a user, obtaining a set of attributes to be selected corresponding to the session message in a main attribute set, wherein the main attribute set is the set of topic types that satisfy the preset conditions in each domain corresponding to the knowledge map; according to the target word, obtaining the confidence degree of each attribute to be selected belonging to the attribute set to be selected; determining the selected attribute corresponding to the maximum value in the confidence degree as the intention corresponding to the conversation message, the intention being the topic type corresponding to the conversation message; through the preset main attribute set, eliminating the calculation of confidence between the non-main attribute set and the target word. The session message processing method and device can quickly identify the intention of the conversation message, also can make the corresponding attribute of the target word more accurate, are more appropriate to the user's intention, and improve the user experience.

Description

technical field [0001] The present invention relates to the technical field of communication, in particular to a method and device for processing session messages. Background technique [0002] In the context of technological development and application requirements, artificial intelligence (AI) has been valued by major companies. The ultimate goal of AI is to build a general dialogue system, and the automatic question answering system, as the most similar form to the dialogue system, has always been the focus of AI research. In the automatic question answering system, the analysis of question intent is a very critical step. Refers to the correct identification of the vocabulary of intent that expresses the need in the question. [0003] In the existing technology, the entity recognition tool of NLP is mainly used to identify the entity in the question, and the entity set in each field is pre-defined by labeling, that is, the intent set; finally, the word or phrase in the q...

AI Technical Summary

Problems solved by technology

[0004] However, this method relies on a large amount of labeling work, the efficiency of intent recognition is low, the response data output for the user is far from the user's question intention, and the user experience is low.

Images