Question and answer prediction model training method and device and question and answer prediction method and device

Abstract

The invention provides a question and answer prediction model training method and device and a question and answer prediction method and device. The training method comprises the steps: acquiring a training sample and a sample result, wherein the training sample comprises a sample problem and sample category information of a sample user; respectively inputting the sample problem and the sample category information into a self-attention layer for processing to generate a first sample problem matrix and a first sample category information matrix; inputting the first sample problem matrix and thefirst sample category information matrix into a mutual attention layer for processing to generate a second sample problem matrix and a second sample category information matrix; performing matching degree calculation processing based on the second sample problem matrix and the second sample category information matrix to obtain a prediction result; and performing calculation based on the prediction result and the sample result, and performing iterative training on the question-answer prediction model. The question and answer prediction model training method provided by the invention is good in training effect and high in model accuracy.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a question and answer prediction model training method and device, a question and answer prediction method and device, computing equipment, and a computer-readable storage medium. Background technique [0002] The question-and-answer prediction system is a system that divides the question text and the user into topic categories, and then calculates the question text category and the topic category that the user is interested in, and infers the probability of the user answering the question after recommending the question to the user. For example, if user A asks the question "Which year was the prose "Back View" created?", and the topic that user A is interested in is "literature", then the above question is divided into the category of "literature", and the topic category that user B is interested in is "Physics", then according to the topic category of the above ...

AI Technical Summary

Problems solved by technology

[0003] The current question and answer prediction system, on the one hand, divides users into categories based on the topics of interest selected by the user when they register and the topics they have answered; To divide the topic category is to cluster the categories of the two entities, ignoring the original text features of the question. For the question text, the clustering features focus on the topic categories of interest to the users who asked the question and the users who have answered the question , while the original text feature focuses on the semantic information of the question text content, but only determines the category of the question it raises based on the category information of the user associated with the question, ignoring the meaning of the question text itself, and the classification of the question is not accurate. , and if a new user who has no category classification asks a new question, that is, a user who does not select a topic of interest raises a question, the question cannot be accurately classified based on the topic category that the user who asked the question is interested in, and then the probability inaccurate forecast

Images