Joint Modeling Of Learners' Emotions And Topics In SPOC Forum

The discourse data in the SPOC forum,as an important carrier of learning feedback,not only reflect the content that the learners are most concerned about,but also hidden individual and collective academic emotions,which provide a reference for exploring the learners' learning states.Discourse data are dynamically changed over time.However,within the textual data,only static content are often used to analyze learning behaviors,the characteristic of the dynamic evolution of topics and emotions is often ignored within the discourse process.To this end,combining the "Freshman Seminar" course's post data in Cloud Classroom platform of Central China Normal University and the final score data of the learners,this study explored the detailed content learners are interested in and the corresponding emotion and its evolutionary trend,analyzed the differences in the emotion-topics and its evolutionary trend of different-achieving groups with the help of the temporal emotion-topic model.So that teachers can timely acquire the psychological state and learning appeals of the learners,and achieve precise teaching interventions.The main contents of this study are as follows:1.Combining the research background of discourse behavior analysis and academic emotion recognition in the field of learning analysis,this study introduced the current research status of topic-emotion mining and its evolutionary trend,and put forward the research content.2.Based on the structure of the classical sentence-level emotion-topic unification model,we added time and emotional prior knowledge(positive,negative and confused)and proposed a more fine-grained and unsupervised temporal emotion-topic model to explore the detailed topics that learners are interested in and the corresponding emotion,and its evolutionary trend during the semester.3.We combined the post data of the "Freshman Seminar" to explore the effectiveness of the model.The results showed that the performance of the temporal emotion-topic model is better than the benchmark model.Although there are many negative and confused emotions in the learning process,the topic information that the learners are interested in reflected their correct attitude to learn English.During the semester,learners have the highest positive emotion attitude towards English at the initial stage.Over time,the negative emotion and confusion of English learning and flipping classroom teaching patterns are the highest.Until the end of the semester,learners' negative emotion remained high.4.In order to explore the relationship between emotion-topic and its evolution trend and learning effect,this study used repeated measures ANOVA to explore the differences in emotion-topic and its evolutionary trends of learners with different-achieving groups.The results show that there is no significant difference in the emotion-topic of different-achieving groups,but there are significant differences in their evolutionary trends.At the beginning of the semester,the high-medium-and low-achieving groups showed positive emotion and recognized the importance of English,but their cognitive levels were different.The medium-achieving group showed higher negative emotion,while the low-achieving group paid more attention to the assessment results.In the mid-stage,the negative and confused emotions of high-medium-and low-achieving groups were dominated.At the end of the period,the positive emotion of the high-achieving groups was mobilized,and the negative and confused emotions of the medium-and low-achieving groups remained high.