A Multi-Category Content Classification Approach Using DistilBERT Transformer and BiLSTM

The exponential growth of digital content has granted individuals unparalleled opportunities to access an extensive reservoir of knowledge. Nonetheless, the increasing presence of hazardous and explicit content inside the continuously expanding digital environment presents notable hazards to persons utilizing the internet, particularly those in younger age groups. The need of implementing effective mitigation techniques is highlighted by the potential harm that can arise from various forms of content, including text, audio, video, and pictures, which may involve violence, explicit material, and other harmful information. In order to achieve this objective, a variety of web filtering methods have been implemented to protect users, specifically minors. These methods include keyword-based filtering, Uniform Resource Locator (URL) filtering, intelligent analysis, and semantic analysis. This work presents a novel model that utilizes the combined capabilities of the DistilBERT Transformer and Bidirectional Long Short-Term Memory (BiL-STM) architecture. This model enables precise classification of website content into a maximum of ten separate categories. The aforementioned categories span a wide range of subjects, including Science and Technology, Food and Recipes, NSFW(Not Safe For Work) content, Entertainment, Politics, Business, News, Fitness and Well-being, and Blogs.