Prototype Design of a Multi-modal AI-Based Web Application for Hateful Content Detection in Social Media Posts

Hate-Speech Detection and filtering of hateful content is an important aspect of any social media post. The ever increasing amount of content posted daily on social media has led to an excessive amount of digital hate being spread in the form of posts, images and comments. The proposed system is developed in order to act as a tool for determining if a particular social media post is hateful and is aimed to aid any benign social media user who has been affected by hate speech and wants to report it. The proposed system uses a multimodal artificial intelligence based approach by classifying different formats of posts, i.e., images and comments or captions separately. An ensemble convolutional neural network architecture is used for this classification, thus, proving to be a strong tool for finding evidence of any prevalent hate speech. This system is tested using the Likert scale for its user interface, accuracy and utility. Based on the result this paper proposes a prototype design of a web application which can be used for hateful content detection.

Recommended citation: Pradhan, T., Bhutkar, G., Pangaonkar, A. (2022). Prototype Design of a Multi-modal AI-Based Web Application for Hateful Content Detection in Social Media Posts. In: Ardito, C., et al. Sense, Feel, Design. INTERACT 2021. Lecture Notes in Computer Science, vol 13198. Springer, Cham. https://doi.org/10.1007/978-3-030-98388-8_36
Download Paper