Research On The Analysis Model Of English Composition Topic Opinion

As a global language,English is one of the important language courses for Chinese students,which has the function of helping students broaden their horizons and increase their knowledge.As an important form of output and the basic skills of second language learners,English composition has received increasing attention from language researchers and educators.However,in the actual teaching process,there is a "one to many" teaching model between teachers and students,and teachers are unable to quickly,frequently,and accurately grade and guide students’ English compositions,resulting in slow improvement in students’ English composition writing.In recent years,with the rapid development of computer networks and artificial intelligence,it has become possible to design and implement an automatic correction system for English compositions.This system can not only alleviate the pressure of manual teachers on correcting compositions,but also improve the speed at which students obtain feedback on English compositions.Currently,among the automatic grading systems for English compositions proposed at home and abroad,there is less discussion on semantic analysis of potential topics and evaluation of emotional perspectives and attitudes in compositions,leading to bias in the analysis of composition content quality in automatic grading systems for English compositions.Therefore,researching and designing a topic opinion analysis model for English compositions has important practical significance.The thesis proposes a topic opinion analysis model for English compositions,which focuses on two parts: thematic semantic analysis and emotional perspective and attitude.The model can automatically obtain the deep semantic information of potential topics in students’ compositions,and obtain the emotional perspective scores based on the topics in their compositions.From the perspective of sentence level,it provides the scores and comments based on the analysis of topic perspectives for English compositions to be approved through text semantic similarity algorithms,objectively and truly reflecting the students’ English writing level.Based on the above research objectives,the main research contents of this article are as follows:(1)This paper studies and designs a contextual topic recognition model that integrates conceptual knowledge(PLDA-NTM).This method uses the conceptual Semantic information of the Probase conceptual knowledge base to introduce conceptual prior knowledge into the traditional LDA probabilistic topic model to generate the global topic semantic distribution of English compositions.Combined with the local semantic representation of English compositions generated by the pre training language model BERT,more accurate topic representation and topic clustering results of English compositions are obtained.(2)A topic based emotional perspective analysis model(TA-GRU)is constructed.This model uses the Bi-GRU neural network structure to obtain semantic information for compositions,and cooperates with the topic level attention mechanism to extract topic emotional perspective information for compositions to be approved,obtaining topic based emotional perspective categories for compositions.(3)Design and implement a composition topic perspective quality analysis module,which calculates the semantic similarity,relevance,and topic based emotional perspective quality scores of the composition from a sentence level perspective,and obtains weighted summation of scores and comments for students’ compositions on topic based sentiment opinions.(4)Several experiments have been conducted to compare the topic opinion analysis model constructed in this article under different English composition corpora.The experiments show that the topic opinion analysis model constructed in this article has higher accuracy and practical value in English automatic scoring systems.