An Empathetic Knowledge-Grounded Conversational System for Mental Health Counseling and Legal Assistance

Context & Background

Victims of harassment and abuse, especially women and children, face massive barriers in seeking legal advice and mental health counseling. AI chatbots can provide a first-line, anonymous, and accessible portal for emotional support and legal awareness.

Problems to be Addressed

Most chatbots lack emotional intelligence, providing robotic responses that can alienate distressed users. They also lack integration with factual databases, leading to hallucinated legal advice.

Aims and Objectives

1. Build a multilingual, empathetic chatbot for counseling and legal help.
2. Train dialogue models on Hinglish (code-mixed Hindi/English) inputs.
3. Integrate a legal knowledge graph to retrieve factual IPC sections.

Methodology

Dialogue act classification, intent detection, and slot-filling are handled by hierarchical deep learning models. Dialogue managers handle conversation states, and generative models incorporate sentiment to output empathetic text. Factual knowledge is injected from curated legal FAQs.

Expected Outcomes

An open-source, pluggable chatbot engine deployed on WhatsApp and Facebook Messenger, trained models for Hinglish sentiment analysis, and research publications in ACL and EMNLP.