Question and answer matching method and device

Abstract

The embodiment of the invention provides a question and answer matching method and device, and the method comprises the steps: carrying out the extraction of a sentence trunk for a current user question, and generating a current sentence trunk corresponding to the current user question; searching a current knowledge point title corresponding to the current sentence trunk from a pre-established mapping table according to the current sentence trunk; wherein the mapping table is established based on historical questions marked as answering errors; if a current knowledge point title correspondingto the current sentence trunk is found, obtaining a current knowledge point; if so, taking the current knowledge point title as the input of a pre-trained general question and answer model, and obtaining a first target answer corresponding to the question of the current user through the output of the general question and answer model, so that errors in the question and answer matching process canbe efficiently repaired.

Description

technical field [0001] One or more embodiments of this specification relate to the computer field, and in particular, to a question-answer matching method and device. Background technique [0002] In the application of intelligent customer service, when customers interact with the customer service robot, it often happens that the customer service robot matches the wrong question-answer pair. On the one hand, user language is colloquial and simplified, and questions with less than 10 characters account for 50%+. This situation is especially obvious in the first questions of customers; on the other hand, ordinary users are not familiar with complex business logic. Complete familiarity leads to unclear, inaccurate, or redundant descriptions of users' problems, and rapid business logic iterations also bring many new user problems to customer service robots. Both of the above two situations may cause the customer service robot to match wrong answers to user questions, so it is n...

AI Technical Summary

Problems solved by technology

On the one hand, user language is colloquial and simplified, and questions with less than 10 characters account for 50%+. This situation is especially obvious in the first questions of customers; on the other hand, ordinary users are not familiar with complex business logic. Complete familiarity leads to unclear, inaccurate, or redundant descriptions of users' problems, and rapid business logic iterations also bring many new user problems to customer service robots

The above two situations may cause the customer service robot to match the wrong answer to the user's question, so it is necessary to repair the error in the question-answer matching process

[0003] Currently, customer service robots are usually based on pre-trained models to match answers to user questions. Model training generally has the characteristics of a long cycle, and the iteration cycle is generally weekly. At the same time, it also needs to go through bypass tests, AB tests and other quality verification links. Now In existing technologies, repairing errors through model training is very inefficient or even infeasible

Images