METHOD AND SYSTEM FOR INTELLIGENTLY RECOGNIZING AND CORRECTING QUESTION

摘要

Disclosed are a method and system for intelligently recognizing and correcting a question, for use in automatically recognizing stems and the answer content of multiple types of questions in a test paper, and automatically correcting the answer content by means of artificial intelligence. The technical solution of the present invention comprises: pre-training, on the basis of a large variety of test paper samples, neural network models (such as a CNN model) into a model for recognizing the question type; using the model obtained by training to detect the position of different questions, the types corresponding to the questions, and different components of the questions (the stem, the answer, and picture) in the test paper; and then respectively recognizing printed character information of the stem and hand written character information of the answer content on the basis of a recognition model; finally, automatically correcting on the basis of the recognized character information.