Machine Learning Techniques for the Automated Identification of Legal Non-Compliance in Online Terms of Service

Due to the complexity of ToSs, it becomes difficult to meet legal requirements for ToSs because of GPA and CCPA among other jurisdictions. For this reason, this research has suggested the development of a machine learning model consisting of both SVM and Bi-directional Long Short-Term Memory (Bi-LSTM) for the identification of ToS non-compliance. The model uses linear and sequential functions for pattern reconstruction, which mitigates the complexity that legal language and the difference between jurisdictions may pose. When tested on multiple jurisdiction data this presented an accuracy of 88.8%, a precision of 91.5% and recall of 87.5<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, an Fl-measure of 89.4, and an AUC ROC value of 0.92. The mutant one proposed method was found to be accurate by 12% more than the other models and was 15% more precise in comparison to other models. This makes it suitable for online use given that compliance can be monitored in real time with feedback. It is apparent from this analysis that the proposed model works well on a legal-noncompliant detection platform providing a strong platform for improved regulation of services offered online.