A test question classification and difficulty classification method based on a recurrent neural network

Abstract

The invention discloses a test question classification and difficulty classification method based on a recurrent neural network. According to the method, a two-stage neural network is used, a first-stage network can classify test questions with unmarked subjects, meanwhile, the classified test questions are put into a second-stage network, and the approximate reference difficulty can be provided for the test questions without marked test question difficulty. In addition, the attributes of the test questions and the difficulty of the test questions have a certain association relationship; for example, the simulation questions of some provinces are often difficult to have other questions, the same questions include years, question types, knowledge points and the like and have a certain incidence relation with the difficulty of the test questions, and therefore feature vectors of the test questions can be constructed through the attributes to serve as training samples for deep learning.

Description

Technical field [0001] The present invention relates to the field of difficulty grading, in particular to a method for classifying test questions and difficulty grading based on a recurrent neural network. Background technique [0002] With the continuous development of computer network technology, computer network testing has increasingly become an important assessment method and method. Supporting online examinations requires a huge examination question bank to facilitate the randomness and objectivity of the online examination system to select and form papers. At the same time, mobile Internet, smart terminal equipment and social network platforms provide a guarantee for the generation, collection and analysis of all-round and massive information, and promote the development and popularization of personalized online education. [0003] However, there are a large number of unclassified and unclassified questions in the Internet question bank. The existing classification method ...

AI Technical Summary

Problems solved by technology

[0003] However, there are a large number of unclassified and ungraded questions in the Internet question bank.

The classification method of the difficulty of existing test questions is often the method of manual grading. However, using manual grading requires a lot of manpower and time, which is not conducive to the promotion of personalized education.

In recent years, with the popularization of deep neural networks, some studies have used neural networks to classify test questions that do not indicate subjects. However, these studies only stop at classifying test subjects and cannot meet the needs of difficulty classification.

Images