In 2011/12, KevTech Apps commissioned a joint research thesis titled “Social Network Behaviour Classification” by David Roberts from the University of Queensland.  The project was supervised and reviewed by Associate Professor Marcus Gallagher (PhD), Professor Janet Wiles (PhD) and Dr Daniel Angus (PhD).

The objective of the project was to determine how effectively machine learningtechniques could be applied to detecting potential instances of derogatory, offensive, sexual, swearing and threatening content on a popular social networking sites such as Facebook and Twitter. The aim was to improve on the accuracy obtained by current commercial solutions relying on comparatively simple matching of key words and phrases.

The project was divided into two phases:

Phase 1:  concentrated on detecting inappropriate content in individual social network posts, in the absence of any contextual information. This compares with the approach taken by existing commercial products.

Phase 2: focused on utilizing contextual information, in the form of entire conversations, to further improve the relevance of detections. The aim of this was to be able to analyze the intent displayed by interacting users, rather than detecting only local language features.

Phase 3: in order to detect cyberbullying and stalking we designed and implemented Patent Pending technology which would detect behaviours over a period of time.

Since this time we have been designing, developing and building a software solution, SafeKidsPro, which detects and report to parents the occurrence of these anti-social behaviours.