Detecting Topics of Chat Discussions in A Computer Supported Collaborative Learning (CSCL) Environment

• Published in: The Turkish online journal of distance education TOJDE Vol. 20; no. 1; pp. 96 - 114
• Main Author: AFACAN ADANIR, Gulgun
• Format: Journal Article
• Language: English
• Published: Anadolu University 01.01.2019
• Subjects: Computer Assisted Instruction; Computer Mediated Communication; Cooperative Learning; Discussion (Teaching Technique); Educational Technology; Foreign Countries; Graduate Students; Group Discussion; Mathematics Instruction; Teamwork; Technology Uses in Education; Turkey
• ISSN: 1302-6488; 1309-4564; 1302-6488
• DOI: 10.17718/tojde.522398

Learning groups' conversations in computer supported collaborative learning (CSCL) environments result in significant information regarding the content of the course. This information is beneficial for instructors to analyze learners' activities during their collaboration process. In understanding these activities and performance of learners, the topic of conversation is important. The purpose of the study is to detect topics of chat discussions conducted by groups of learners while collaboratively studying in an online CSCL environment called Virtual Math Teams (VMT). We implemented the study in the context of a graduate level course during one term in a large state university in Turkey. Participants are MSc and PhD students registered to the course and divided over five groups of three students. We combined and employed methods of data mining, social network analysis, and topic detection to identify topics of learners' discussions. Our data analysis process aims to identify the task related topics occurred in chat discussion of learning teams. In our analysis we followed the stages of data preprocessing, segmentation analysis, and topic detection. Our purpose with the preprocessing stage was eliminating improper data for the main analysis and making the data ready for analysis stage. Therefore, our final corpus was shaped to involve 95% of initial chat messages. Segmentation analysis aims to explore organization of chat discussion and divides the chat logs into more manageable units according to their corresponding contents. In total, we resulted 294 segments including task related and non-task related ones. The topic detection analysis explored the content of chat segments and revealed the major subject of discussions with the use of latent semantic analysis, which is applied to find content similarity among segments and indicative words produced through the use of two mode network analysis.