Online Nepali Abusive Text Detection for Intimate Partner Violence Research
This project enables large-scale detection and analysis of online Nepali abusive text to support research on intimate partner violence (IPV), providing ways to understand prevalence and context better while informing interventions.
The increasing virtual social engagement has led to a rise in online abuse and violence, specifically online Intimate Partner Violence (IPV), whose nature and prevalence are not well known. A key challenge is the ability to automatically detect abusive texts and contexts in the Nepali language, hindered by a gap in good Nepali NLP models and datasets.
Our goal is to build AI-powered systems that detect and classify abusive Nepali text, supporting IPV research and prevalence estimation. We achieve this by creating annotated datasets from social media posts, training models to identify various forms of abuse, and developing web-based platforms for scalable text analysis.
This project has developed a chat application for simulating IPV-related and normal conversations, created comprehensive annotated Nepali Twitter datasets to identify abusive content, and built a sophisticated web platform that collects social media text, classifies abuse types, highlights key phrases, and allows human review and correction for accurate analysis.