A Study On Construction And Application Of Online Learner Aspect-Opinion Mining Model Based On The Method Of NLP

With the rapid development of the Internet,online learning has become an important form of learning and the scale of online education is increasing day by day.Learners using online education platforms generate large amounts of text,and artificial mining is time-consuming and energy-consuming with low efficiency.These learners are constantly producing content,mainly in the form of text.Manually analyzing a large number of text generated by learners is mostly a one-off task.Because of the large number of course learners,it is difficult for any team to accurately analyze the relevant information of learners' opinions,learning situation,resource preferences,etc.Therefore,with the development of artificial intelligence,natural language processing and text mining technology,mining the opinions in learner-generated text is conducive to understanding learners'evaluation of teachers and resources,mastering learners' learning situation,recommending personalized resources,conducting learning analysis,and constructing learners' opinion knowledge base,which is conducive to aggregation of collective wisdom.The research revolves around three research questions:first,identify learners' viewpoint sentences and extract opinion elements.Secondly,construct a learner's Attribute-level view mining model.Thirdly,analyze the teacher and resource evaluation with the results of model extraction.The study sorts out the related research status through literature review;recognizes and compresses opinion sentences based on rules to abstract learners' opinions with natural language processing method;designs learner opinion five-tuple elements extraction algorithm to structurize unstructured text information;analyzes dependency syntax and lexical semantic association to obtain attribute-opinion word distribution;uses mathematical modeling method to build,verify and optimize the learner-aspect opinion mining(LAOM)model integrated with semantic relationship set and word distribution;analyzes teacher's teaching behavior and resource quality using the results extracted through the LAOM model.The research work of the LAOM model mainly includes the following four parts:A method of opinion recognition in text data generated by learners.Coarse-grained text mining is aimed at document-level,text-level and paragraph-level text data processing,where learner-generated tex t is mainly short-text oral expression and not every sentence contains learners's opinion information.Fine-grained mining method,which aims at sentence-level,lexical level and attribute-level,is more suitable for learner opinion mining.Attribute-level learner opinion mining is to extract and use the viewpoint elements in opinion sentences.It is based on grammatical rules to recognize and compress the opinion sentences.It combines semantic meaning to recognize and compress the opinion sentences from the text generated by learners.It can improve the accuracy of learner opinion mining,reduce the amount of data calculation and improve the efficiency of viewpoint elements extraction.The LAOM model incorporating semantic relations.In attribute-level learner opinion mining,it is helpful to obtain low-frequency attributive words and opinion words by combining the semantic relations between words.The LAOM model acquires the Must-links and Cannot-links semantic relationships among attributes,opinion words,attributes and opinion words,and integrates the semantic association set and non-association set between attributes,opinion words,attributes and opinion words into the model to enhance the extraction accuracy and recall rates of low-frequency attributes and opinion words.The prototype system of the LAOM model verifies the accuracy and validity of the five-tuple opinion extraction from learner-generated text.Analysis of teacher's teaching behavior in learners' opinions with the LAOM model.In online learning,teachers' teaching behavior is an important factor influencing online learners' course evaluation.This study proposes the research hypothesis that teacher's teaching behavior is related to learners' course evaluation.The study conducts a statistical analysis of learners' opinions extracted through the LAOM model and an in-depth mining and analysis of learners' opinions related to teacher's teaching behavior.It adopts the SAT coding table of teacher's teaching behavior,classifies and codes opinion records related to teacher's teaching behavior in learners' evaluation information,verifies the coding reliability and validity,and then analyzes the correlation between different teacher's teaching behaviors and learners' course evaluation.Analysis of the resource quality evaluation in learners' opinions with the LAOM model.Online educational resources are an important part of online learning,and learners' opinions contain the related viewpoints of educational resource quality evaluation.The study proposes the hypothesis that the quality of educational resources is related to learners' course evaluation.It conducts a statistical analysis of learner's opinions extracted through the LAOM model,and mines and analyzes learners'opinions related to the quality of educational resources.It classifies and codes opinion records related to educational resource quality in learners' evaluation information by PEI coding table,verifies the coding reliability and validity,and analyzes the correlation between the evaluation of various aspects of educational resource quality and learners' course evaluation.In view of the limitation of traditional opinion mining model lacking semantic association,this study proposes a rule-based LAOM model for fine-grained attribute-level learner opinion mining by recognizing the sentences containing subject-predicate or neutral relations in learner-generated text as opinion sentences and combining the lexical semantic relations between attributive words and opinion words.Compared with SRC-LDA,ACM,LDA and CAMEL through experiments,the LAOM model extracts low-frequency attributes and opinion words with higher accuracy and recall rates.According to the LAOM model,an opinion mining prototype for batch processing of learners' text data is designed,and the learners' opinion extraction results are applied to evaluation of teacher's teaching behavior and educational resource quality.The results show that the teacher's teaching assistant behavior(evaluation feedback,technical guidance)is weakly related to the learner's curriculum evaluation.Teacher management behavior(organizational discussion,curriculum management,assistant learning resource management)is positively correlated with learner curriculum evaluation.Teacher bishop behavior(instruction,instruction,presentation)is positively correlated with learners'curriculum evaluation.The satisfaction of resources to learners'individual factors is the primary factor for users to evaluate resources;learners' individual factors and network learning environment are positively correlated with resource evaluation;learners emphasize the satisfaction of resources to their individual factors in network learning space;online education resources have an impact on the support and accessibility of different hardware devices.Learners evaluate the quality of resources.Combining learners' opinion information with other types of data,and applying the research outcomes to personalized recommendation of learning resources and customizing personalized learning path will be the important directions of future research.