Multi-round intelligent question answering method, system, controller and medium

Abstract

The present invention relates to a multi-round intelligent question answering method, system, controller and medium. The method includes: acquiring question sentence information of the current round and previous round; Round's intents and entities, and previous rounds' intents, entities, and answer template types; select the optimal answer based on the obtained current round's intents and entities, and previous rounds' intents, entities, and answer template types. The present invention realizes multiple rounds of question and answer based on intention recognition and entity extraction, improves the accuracy of multiple rounds of question and answer, and reduces maintenance cost.

Description

technical field [0001] The present invention relates to the technical field of human-computer interaction, in particular to a multi-round intelligent question answering method, system, controller and medium. Background technique [0002] At present, the intelligent question answering system is mostly constructed in the way of one question and one answer. Some multi-round question and answer systems simply solve the interaction problem based on the post-processing judgment of intent entities, and cannot effectively combine the information left by the current question and answer questions to process multiple rounds of question and answer. Machine responses to questions and answers. [0003] However, for more and more task-based question and answer needs, it is often necessary to save the information left by the human-computer interaction before and after, and based on the past interaction information, the machine throws questions to guide customers to complete the task-based q...

AI Technical Summary

Problems solved by technology

[0002] Nowadays, intelligent question answering systems are mostly constructed in a question-and-answer manner. Some multi-round question answering systems only solve interactive problems based on post-processing to determine the intent entity, and cannot effectively combine the information left by the current question and answer between questions and answers to process multiple rounds. Machine Reply for Questions and Answers

[0003] However, for more and more task-based Q&A needs, it is often necessary to save the information left by the human-computer interaction before and after, and based on the previous interaction information, the machine throws questions to guide customers to complete the task-based Q&A, which requires relatively high requirements for the Q&A system. The previous simple interaction and rule-based multi-rounds cannot meet today's task-based multi-round question answering, with limited expressive ability and difficult maintenance

Images