S2-O/LT5-3 - Automated Discussion Analysis - Framework for Knowledge Analysis from Class Discussions

3. Research Full Paper
Swapna Gottipati1 , Venky Shankararaman1, Mallika NITIN Gokarn1
1 Singapore Management University

This full paper presents our research work on the application of automated classroom discussion analysis techniques for enhancing student learning in computing education.

Learning is an interactive process between the student, the teacher, and the subject matter. Learning is enabled by various teaching approaches that motivate students’ desire to learn and empowers students to think about the subject matter on their own. Classroom discussion is a sustained exchange between teachers and their students with the purpose of developing students’ capabilities or skills on a specific topic. It provides a cooperative learning method that promotes students to interact with other students and the instructor. According to the model illustrated by the “learning pyramid” developed at the National Training Laboratory (NTL), when used as a teaching approach, classroom discussions is an active study method that can lead to greater retention of information and material studied, and higher academic achievement. Classroom discussions provide opportunity for effective personalized and collaborative learning across various education programs. In-class discussions, as well as online discussion forums should be carefully designed and executed by the instructors to stimulate student thinking, and increase participation and engagement.  

Analysing the discussions helps instructors gain better insights on the personal and collaborative learning behaviour of students. Thus providing directions for making appropriate changes to the content and delivery so as to enhance student learning behaviour. For example, by analysing the individual student participation in the discussion forum enables the instructor to provide participation grade and necessary feedback for effective personalised learning; by analysing the topics that the individual student focusses upon will enable the instructor to discover the strengths and weakness of each student with regard to specific topics in the course. Additionally, discussions provide a rich source of content knowledge that can help students enhance their problem solving and collaborative learning skills. For example, the posts on specific topics can be summarised and shared with all the students in the cohort, thus supporting topical revisions or improving the problem solving skills for projects or assignments. Based on literature review, we understand that most of the online discussion forum data is not effectively used due to the fact that the manual process of generating insights and high quality information is a very tedious task. Additionally, knowledge from in-class discussions is not effectively captured and mined due to lack of appropriate automated tools.

In this paper, we propose a conceptual framework, Automated Discussion Analysis (ADA), as shown in Figure 1 for analysing student discussions. This framework provides a starting point for the community of stakeholders to consider how discussion analysis can help in making informed decisions with respect to teaching, learning, and course improvements. It elaborates on the main components of a classroom discussion analysis model and the interactions between these components. It also illustrates the main aspects to be taken into account when implementing discussion analysis tools using data mining, text analytics, and natural language processing techniques. According to Woolf’s AI grand challenges in education, “Interaction Data to Support Learning’ is the third AI challenge that will help address the goals of education. Our framework supports the vision of the challenge –“adaptable to acquiring and analysing educational data and discovery of novel and potentially useful information”.  We therefore, present the relevant AI based education tools based on this framework. Furthermore, in this paper, we present a case study where the framework is applied to a course in our school.

Figure 1: Conceptual Framework for Classroom Discussion Analysis

The paper is structured as follows: In Section 1, we review other related work in two areas namely class discussions pedagogy, and analysis of discussions in computing education. Section 2 presents the limitations of current approaches and need for a unified framework to better understand the various analysis that can be performed on discussion posts. In Section 3, we present the Classroom Discussion Analysis Framework along with detailed description of the various components of this framework, how these different components interact with each other, and also highlight the key aspects to be taken into consideration when implementing discussion analysis tools. In Section 4, we describe the AI education systems based on this framework. In Section 5, using a case study we present where and how the framework is applied to a selection of courses in our school. Section 6 concludes with a summary and discussion on the usefulness of the framework and areas for future work.