Research On Automatic Scoring Model Of Chinese-English Interpretation Based On Semantic Scoring

Intelligent education is the major trend of education development in the Internet age,and computers have become an important tool for assisting teaching.In oral English teaching and automatic scoring,the automatic scoring system for questions that specify candidates’ answers,such as reading aloud questions and imitate reading aloud questions,has reached a practical level in recent years,while the research on automatic scoring methods for questions with a given answer range,such as retelling questions and interpreting questions,is in its infancy.And there are few related researches.Interpretation test is a comprehensive examination of foreign language application ability,including language thinking ability and language organization ability.The research and development of an effective Chinese-English automatic scoring system can not only provide students with an interpretation practice platform,but also assist teachers in teaching and release teachers’ pressure on teaching and marking papers.Based on this,we focus on the semantic scoring method at the interpretation content level by analyzing the scoring requirements of interpretation test,and construct a multiparameter automatic scoring model for Chinese-English interpretation in sentence-level,as the foundation of building an application system.This study focuses on semantic scoring of oral Chinese-English translation,and introduces a model which integrates long-term memory neural network and selfattention mechanism.The model can be applied to keyword scoring and sentence semantic scoring.The scoring principle of the model is as follow: firstly,the features of words and sentences are extracted and expressed in vectorized form;secondly,the feature vectors are optimized by bidirectional short-term memory neural network;thirdly,the semantic features of words or sentences are obtained by self-attention mechanism;finally,the semantic scores are calculated by a simple neural network.Experiment shows that compared with the semantic scoring model based on unfolding recursive autoencoders,which performs well in semantic scoring,this model has better results in sentence semantic scoring.The average coincidence rate between the scoring result of this model and the original score is 55%.Based on the study of semantic scoring,we proposed a multi-parameter automatic scoring model for Chinese-English interpretation in sentence-level.The automatic scoring model based on semantic scoring calculates the scores of candidates by fusing the characteristic parameters of phonetic level and content level.At the phonetic level,fluency is selected as the characteristic scoring parameter,and the model directly scores the fluency characteristic of the candidates’ recording.At the content level,two feature scoring parameters,keyword and sentence semantics,are selected to score the content features after the candidates’ recording is converted into text by manual conversion.Specifically,the zero energy product method is used to extract speech features and calculate fluency feature scores.The semantic scoring model introduced in this paper is used to calculate the semantic feature scores of keywords and sentences.Finally,the random forest algorithm is used to fuse the above three feature scoring parameters to obtain the total quality score of Chinese-English interpretation.Through the performance test of the automatic scoring model based on real test data,the average coincidence rate between the scoring results of the Chinese-English interpretation automatic scoring model proposed in this paper and the original scores reaches 77%,which verifies the validity of the characteristic scoring parameters selected in this paper and proves the feasibility of the Chinese-English interpretation automatic scoring method based on semantic scoring studied in this paper.