Research On Topic Mining In MOOC Discussion Board

It is everyone's concern that education shall embrace the modern age of information.With the trend of educational big data,a growing demand of informatics aided teaching-learning have been raised.Not only high-quality content is needed for MOOC teaching,analytical procedural and summarized result of data-driven teaching is also required so that the result will yield as the bond of student-teacher and student-student interactions.It would also urge the teacher to effectively monitor the course activities,to adjust teaching strategies accordingly,and to guide the learners in a correct learning path.These results are also direct feedback to learners facilitating them to engaging in learning activities with more agency and self-consciousness.During the teaching process of MOOC,text-data in discussion are important showcase of on-course interactions as well as communications and collaborations.Moreover,it also serves as important data evidence of learning progress and reviewing teaching qualities.Every second on MOOC platform a formidable amount of text-data was created and it is considered impossible to carry out the statistics via manual reading and tagging.It is also noticeable that great similarities was shared in text-semantics between learners and topics will vary with the flow of time.It could be valuable to automatically and efficiently data mine and demonstrate hidden semantics within texts.Based on the course of Modern Educational Technologies,Shaanxi Normal University,Chinese University MOOC Platform,this research focused on the analysis of text of learners' topics and their evolution in discussion.Methods of datamining of MOOC leaners' topics and relative techniques was confirmed during literature review.A pre-procedure was carried out with discussion texts within the learning experience and reflections from each unit in Modern Educational Technologies.Such text specialization methods as text filtering,Chinese words separation based on jieba,Stop-words removal and TF-IDF algorithm was utilized so that text data would be converted into specific vector form that could be processed with topic models.A comparison of categorizing topics and clustering techniques found out the advantage of topic model in processing implicit semantic information.ATM model and DTM model were chosen as the core technique of this research to explore such questions as the content of learners' discussion topics,recommending authors of similar topics and evolution of topic as time flows.Moreover,a visualization was carried out for the result of this research.With the specialty of data result in consideration,the text data visualization could be dichotomized into text-content-based and topic-model-based.Bar chart,word cloud map,heat map,pyLDAvis and t-SNE were used as means of visualization.Datamining topics are a beneficial trial of MOOC's integration of connectivism theories.Additionally,it also provided aid to enhancing student-teacher interactions and building learning communities in open online courses.It is revealed in experiment result that not all units could find its match with topic datamining results.The main concentration of learner's discussion is about the content in the beginning of the course.Also,it is worth mentioning that no significant topic was found in discussions in Unit 8 and 9.Based on the completion of the feature of recommendation of similar topic creators,learners could navigate to similar learning group with ease and efficiency.Plagiarism activities was under surveillance to some extent thanks to this.The topic "Teaching Requirements in the Information Age" reflects the change of the concept of "learners in the Information Age".The improvement of information technology ability is the teaching requirement in the Information Age.The shift in probabilities in keywords within topics indicates the changes of key discussion points.For example,learners started discussing learning theories in the beginning of the topic "Learning Theoretical Basis" and ended in focusing on informational technologies.Core Keywords are the preferred way for students to choose their discussion contents and tend to accompany them with connection words.For instance,students tend to combine cognitivism,behaviorism,constructivism and humanism together,demonstrating a throughout understanding with the topic of learning theories.The most discussed topic in panel Latest Progress in Theoretical Researches is Multiple Intelligence,while keywords in other domains receiving less attention.A strong correlation was found between the frequency of topical keyword usage and the time of a published unit,and some student could keep up with the set pace of learning.Another find is that positive comments and feedback become high frequency words in the middle and end of an overall course session,indicating a progressing procedure was required for learners to approve a course.The heat of each topic shows a steady or upgoing trend.Stability and continuality were found within topic discussions in course learning.The feedback from visualization states that the frequency of chosen topics are similar in size thus no absolute "most heated topic"was found.It is also worth mentioning that lots of discussion text that cannot be identified into a specific topic exists,these learners usually have a broad spectrum of topics.Learners could clearly figure out relevance level of their and mainstream topics.These analytic results could help teachers with a clairvoyance of ongoing course progress so that action could be taken to evaluation and modify course content and structures.Learners could discover learning communities with ease and in-depth discussion could be carried out with common topics.This will also provide reference to actualizing self-adaptive and personalized learning in MOOC courses.