An intelligent scoring method for subjective questions

Abstract

An intelligent marking method for subjective questions, including preprocessing the candidate's answer text and the reference answer text to obtain the corresponding word vector sequence representation; using the semantic feature extraction network to extract the semantics of the sentences in the candidate's answer and the reference answer Vector; use the fusion attention mechanism to calculate the semantic matching degree of the examinee's answer and the reference answer sentence vector; use this matching result to calculate the weighted sentence vector of the examinee's answer; the sequence of sentence vectors for the complete examinee's answer and the weighted sentence vector based on the fusion attention mechanism sequence; use the semantic feature extraction network to calculate the semantic vector of the complete answer and the semantic vector based on the attention of the reference answer; glue the two vectors to form the final vector representation of the candidate's answer. The final vector representation of the examinee's answer is scored using a multi-layer feed-forward neural network. By using the invention, automatic marking of subjective questions can be realized, and the efficiency of marking can be greatly improved.

Description

technical field [0001] The invention relates to an intelligent marking method for subjective questions, belonging to the technical field of natural language processing. Background technique [0002] With the development of information technology and the advancement of paperless examinations, many authoritative qualification examinations require candidates to answer on computers. The traditional marking method mainly relies on a large number of trained marking personnel for manual marking, which has high economic costs and low efficiency, and also affects the stability of marking quality due to factors such as subjective differences and physical fatigue of the marking personnel. Automatic machine marking can not only save economic and labor costs, improve the efficiency of marking, but also assist in the supervision of the marking process and improve the overall quality of marking. With the development of natural language processing and artificial intelligence technology, ma...

AI Technical Summary

Problems solved by technology

However, it is still impossible to analyze the answer semantically and logically from the overall text, and it is impossible to solve the situation that some entries match some entries, but the overall logic of the text is incorrect.

[0006] Chinese patent CN108734153A disclosed on 2018-11-02 a method and system for high-efficiency computer marking, which divides candidates’ answers into several categories through synonyms and synonyms matching, and the marking teacher scores each category to reduce the pressure of marking. It is suitable for filling in the blank questions, translation questions and other questions with few answer changes and short texts. In theory, this method also belongs to the scope of simple word form matching, and it is difficult to apply to long answer text reviews

[0007] To sum up, although the existing marking system has achieved the automation of part of the marking process, it still stays at the basic level of word form matching when it comes to the most core subjective question answers, and does not go deep into the text semantics, especially Faced with subjective questions with long answer texts and varied answers, it still needs to be manually reviewed by professionally trained marking personnel

Images