Automated Detection of Social Roles in Online Communities using Deep Learning

Online communities are an increasingly important aspect in the digital age, for business organizations, diverse industry sectors and overall, in modern society. The social role of each end-user, influencers to followers, and content providers to receivers is a primary consideration when evaluating the purpose and contribution of any online community. Most existing research on the detection of social roles in online communities is based on manual observations and analysis. This paper introduces a technique for automating the detection and extraction of social roles from online communities. Given the large volume of text and value of content, it is no longer viable to manually encode and detect social roles and contributions. The machine learning approach is based on a deep recurrent neural network and a word embedding model. A dataset consisting of over 1.2 million textual posts extracted from an online community on higher education in Australia was used to demonstrate the technique. This technique can be applied to any online community to automatically identify social roles, their influence and interactions.