A Study on Keyword Analytics as a Precursor to Machine Learning to Evaluate Radicalisation on Social Media

摘要

Social Media as a platform has evolved from its origins as a digital networking and social communications platform, to be an integrated digital limbic system for system; not only providing direct access to the user, but to gauge and shape societal trends globally. The platforms integrating into a user's digital life and footpath therefore generates large quantities of data about users, and their interactions on and between platforms. The continual advancement and integration of technology and its availability and its availability to wider reaches of society are synergic to the growth, use and popularity of social media platforms; leading to social media becoming an integral flagstone of social communication today. This paper discusses a study conducted on a keyword analysis of data mined from open source social media, in this instance, Twitter, and comparing the results with that of an analysis of data captured from 2 publications from a known extremist organisation. Comparatively, the results of the analyses show that it is difficult to attribute behavioural trends through specific keyword usage. The results from the keyword analysis of the Twitter datasets show that the language and themes used are relative to current “trending” events, whereas the recurring keywords found in the extremist publications are that of religious ideology, not specifically relating to a date or event. The study also shows that context plays an important role in the identification of behavioural trends and radicalisation, and as such, further research utilising a keywords-in-context approach of analysis would glean a richer result set to lay a potential foundation for a machine learning approach to big social media dataset analysis.