Network Community Analysis Based Enhancement of Online Discussion Forums

Enhancement of online learning tools has become essential in the current pandemic. Chat forums involving multiple experts and topics are typically set up online to aid student learning. Effective communication in these forums must be facilitated. In this work, we present a platform-agnostic technique that successfully (a) Identified the relevant topics present in a discussion forum through keyword extraction and tagged each message to these different topics, thus forming conversation graphs for each topic; (b) Used Betweenness Centrality [1] to calculate the leaders in each of the topic graphs for a given time interval, and used an exponentially decaying ranking factor to generate the top-k leaders in each of the topics; (c) Found the topic-wise, expert-wise ideal timing to ask questions on the forum by generating a histogram of availability for each user for each topic. Further, we integrated all the steps in an easy-to-use Python interface. Our work is fully functional and is currently being used at BITS Pilani Goa. We have done performed our analysis on two diverse chat logs. First, Internet Relay Chat Logs of Ubuntu 1, which is a collaborative discussion forum. Second, on a dataset that we have collected at our University. This was a large forum for a course that had been taught in an online mode on the Telegram platform. Our analysis on the University Dataset identified the teaching assistants assigned with 100% recall for all k values > 3. Our work requires no training, and is scalable both in terms of users, and messages. The step that takes the longest time is (b), and it takes 72.1 seconds for an Ubuntu IRC chat with 541776 messages. The activity histogram of an expert ranked 1 in the topic ‘ubuntu’ on the Ubuntu IRC dataset is shown in Fig. 1. Fig. 2 shows topics extracted from the university dataset. Activity histogram of a user ranked 1 for the keyword ‘strings’ in the University dataset is shown in Fig. 3. We find that our tool can help make the transition from in-person instruction to remote, online learning easier.