Method, system and non-transitory computer-readable recording medium for generating customized learning report based on test paper image using artificial intelligence

Abstract

An AI-based learning reporting system for generating a user-customized learning report based on a test paper image according to an embodiment of the present invention comprises: a scoring unit that preprocesses an image uploaded by a user to extract text data and analyzes the extracted text data to determine the correctness of a question; an explanation unit that provides a question explanation based on the result of determining the correctness; and a report generating unit that generates a report based on the question explanation. The scoring unit inputs the text data into an AI scoring model to classify the question by subject unit and subject element, maps and outputs a first achievement level of the user based on the type of incorrect answer, accumulates and records the output first achievement level to calculate a learning curve of the user by subject unit and subject element, and outputs a second achievement level of the user based on the learning curve. The scoring unit outputs the first achievement level based on the following mathematical formula, wherein S1 represents the first achievement level, C is the number of questions determined to be correct among the user's responses, N is the total number of questions, E is the number of questions determined to be incorrect among the user's responses, and R is a specific type of incorrect answer. The number of repetitions, d1, d2, and d3 may represent difficulty adjustment coefficients, and the explanation unit generates a first problem explanation that reflects the instructor-specific solution style using an artificial intelligence problem explanation model learned based on explanation data including textbook explanations, instructor's lecture scripts, and lecture images, inputs the second achievement level and the explanation data into an artificial intelligence explanation difficulty model, and outputs a second problem explanation with adjusted difficulty reflecting a step-by-step solution procedure according to the second achievement level, and the explanation unit calculates an explanation suitability score based on the characteristics of the style condition extracted from the explanation data and the generated problem explanation, and may confirm the problem explanation as the first problem explanation only if the explanation suitability score is greater than or equal to the standard explanation suitability.

Images