A method and device for training robots to imitate and learn artificial customer service

Abstract

The invention discloses a method and device for training robots to imitate and learn manual customer service, introduces a training seat mechanism, and flattens and unifies robot pre-training, intelligent customer service service and update maintenance supervision and learning, optimizes service online and training and learning processes, and goes online immediately , the engine will filter enough confident questions to answer; on the one hand, customers get timely and accurate message feedback, and the service scope covers services other than the engine, seamlessly connecting with engine services; on the other hand, the judgment of the training agent will be based on " "Supervised learning samples" are written into the engine feature analyzer. Sample accumulation, training and intelligent services go hand in hand, which greatly saves the time and cost of various online. With the continuous self-training and learning of the engine, more and more problems can be filtered. The supervised learning workload for training agents will decrease day by day.

Description

technical field [0001] The present disclosure relates to the technical field of robots, in particular to a method and device for training robots to imitate and learn manual customer service. Background technique [0002] Customer service robot applications are a branch of natural language processing. The development of customer service robots is accompanied by the innovation of natural language processing technology. There are roughly four types: the first stage is a "retrieval robot" based on keyword matching; the second stage is to use certain templates to support multiple word matching, It also has fuzzy query capabilities; the third stage is to introduce search technology based on keyword matching, and to sort according to text relevance; the fourth stage is based on machine learning, applying deep learning to understand the intention of intelligent customer service technology. Modern intelligent customer service robots are developed from traditional natural language pr...

AI Technical Summary

Problems solved by technology

Under this solution, the following problems may occur: (1) The keywords entered by the user are "synonymous and different" from the keywords in the preset QA library, and effective recommendations cannot be made; (2) QA keywords for the same keyword There may be too many, and users need to perform more complicated secondary screening to find the answer they expect;

[0009] In the second solution, for the same intention, the expression (sentence syntax) can vary greatly, which means that the early sentence syntax induction and sentence regularization will be very heavy, the cost of modifying operation and maintenance is very large, and professionals are required perform maintenance operations;

[0010] In the third solution, the intelligent customer service system based on machine learning requires a large amount of sample accumulation and learning and training costs in the early stage, and takes a long time to go online. Unpredictable error response in

Images