Similarity evaluation method, answer search method, device, equipment and medium

Abstract

The present disclosure relates to a similarity evaluation method, an answer search method, a device, a device, and a medium, wherein the similarity evaluation method includes: acquiring a target character string and a target text image whose similarity is to be evaluated; inputting the target character string and the target text image To the pre-trained similarity evaluation model; the character feature information of the target string and the image feature information of the target text image are respectively extracted through the similarity evaluation model, and the target string and the target text image are evaluated based on the character feature information and image feature information The content similarity of the character feature information includes the positional relationship and semantic relationship between character features; the image feature information includes the positional relationship and semantic relationship between image features. The above methods comprehensively improve the accuracy of similarity evaluation, help to further improve the accuracy of answer search, and can be better applied to photo judgment.

Description

technical field [0001] The present disclosure relates to the technical field of artificial intelligence, and in particular to a similarity evaluation method, an answer search method, a device, a device and a medium. Background technique [0002] Photo judgment is an important application of artificial intelligence technology in the field of education, which can save teachers' judgment costs and improve the efficiency of judgment. Specifically, the user takes a photo after answering the question, and then uploads the answer image obtained by taking the photo to the application program that can take a photo to judge the question, and the application program recognizes and scores the answer image. [0003] However, the current photo-judgment technology is very limited in application occasions, and most of them can only deal with logically correctable elementary school math questions such as horizontal and vertical, but cannot deal with common question types with semantic inform...

AI Technical Summary

Problems solved by technology

[0003] However, the current photo-judgment technology is very limited in application occasions, and most of them can only deal with logically correctable elementary school math questions such as horizontal and vertical, but cannot deal with common question types with semantic information. The main difficulty The reason is that the current technology is difficult to accurately search the corresponding correct answer from the question bank based on the image of the question type uploaded by the user with semantic information. The inventor found through research that the key bottleneck to break through this difficulty lies in the poor similarity evaluation method. The current similarity evaluation methods between text images and character strings have the problem of low accuracy, and it is difficult to be better applied to the occasion of answer search for semantic information questions.

Images