Question answer pair expansion method, device and equipment and computer readable storage medium

Abstract

The invention provides a question and answer pair expansion method. The method comprises the steps of collecting the questions in a system log; acquiring the questions and answers in the network; establishing indexes for questions and answers in the network; matching the questions in the system log with the questions and answers in the network after index establishment to obtain a predetermined number of candidate question and answer pairs; and expanding the question answer pairs based on the candidate question answer pairs. The invention further provides a question and answer pair expansion device, a computer device and a computer readable storage medium.

Description

technical field [0001] The present disclosure relates to a question answer pair expansion method, device, computer equipment and computer-readable storage medium. Background technique [0002] In human-computer interaction systems such as language assistant systems, artificially created questions and answers are very important resources, because this kind of data is the real reply of human beings, and the quality of the answers is higher than that obtained through social networking sites. The process of getting answers from the website is roughly crawling the real questions and replies of users on social networking sites as questions and answers, but this kind of answers are common because they are automatically crawled and generated in a certain context. The problem is that the answer doesn't match the question. [0003] The artificially created question answer pairs are generally written manually by the labeling team or product managers. This method is not only slow but a...

AI Technical Summary

Problems solved by technology

[0003] The artificially created question answer pairs are generally written manually by the labeling team or product managers. This method is not only slow but also expensive, and is not suitable for large-scale expansion of question answer pairs.

[0004] Although many of the answers obtained through social networking sites do not match the questions, high-quality answers to questions can be obtained through manual filtering one by one. However, because there are too many data automatically crawled through the network, if each Manual filtering takes a very low time-to-benefit ratio, so this method can also be regarded as slow and costly

Images