FP7 Logo

EC Project 257859

European Union

Co-funded by the 
European Union

Understanding the Driving Factors of Discussion Activity on Community Forums

Online communities have now become the de facto resource for users to ask questions, share opinions and debate and interact with their peers in a social environment. Information contained with such communities is useful to market analysts and brand managers as it informs public opinion towards their products and/or services.

By tracking the posts within online communities that turn into lengthy debates analysts are empowered with early and up-to-the-minute insights into public opinion and consensus. In order for such tracking to be effective we need to understand what factors drive lengthy discussions in online communities so that posts which evolve into lengthy debates can be detected early, prior to the debate occurring.

In a recent article published at the Third IEEE Conference on Social Computing entitled 'Anticipating Discussion Activity on Community Forums' [1], authored by Matthew Rowe, Sofia Angeletou and Harith Alani, the authors presented a two-stage approach for predicting discussion activity on the Irish community message board Boards.ie that: first identified seed posts - i.e. thread starters that yield a reply, and second predicts the level of discussion activity that seed posts will yield.

Three feature sets were explored to identify the key factors that drive discussions including properties of the post author - i.e. how they are connected into the community and their participation, properties of the post itself - i.e. quality measures of the post's language such as its informativeness and readability and also the sentiment of the post, and the topical concentration of the author - i.e. how distributed the author's past posts are across the various forums in the community.

The authors found that seed posts - i.e. thread starters that yield a reply - contained fewer hyperlinks than non-seeds - as shown in the above box plots where 'neg' refers to non-seed posts and 'pos' refers to seed posts. The above figure also shows that seed post authors were found to have a greater affinity with the forum in which they were posting than authors of non-seeds. This means that authors who start a new thread in a forum that they do not normally post in, and therefore have less affinity with, are less likely to seed a discussion.

Experiment findings also showed that heightened discussion activity was associated with an increase in the forum likelihood of the author, and therefore the affinity that the user has with the forum they are posting in. Another topical concentration measure is forum entropy that gauges how dispersed a user's past posts are across various forums. The authors found that a decrease in forum entropy, and therefore a more concentrated post history for the seed post author, was associated with increased discussion activity. Content qualities were found to play an important role in predicting discussion activity, in particular, experiments showed that a reduction in the post's sentiment was correlated with an increase in discussion activity - i.e. posts that are more negative will generate more activity.

These findings enable market analysts and brand managers to track posts that will yield a reply and lead to lengthy discussions. The identified factors in driving discussions also inform content creators in shaping their content and behaviour in the community to maximise the impact of the content they share with the community.




[1] Anticipating Discussion Activity on Community Forums. M Rowe, S Angeletou and H Alani. The Third IEEE International Conference on Social Computing. Boston, USA. (2011)


Comments (0)