GrievEase: A Multimodal Deep Learning Framework for Smart and Transparent Petition Management

View at publisher Summarise with VIRAT
Authors: L. Sindhu, V. Mohanraj, Yuvaprasanth L, Praveen Raj M A, Sadhasivam Arumugam, Sri Rama Krishnan B
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

GrievEase is a petition system that simplifies filing grievances in India by enabling a system that has a structured and easy to use interface. It automatically classifies complaints departmentally, by type and urgency with a hierarchical multitask BERT model with each layer briefing the other. Cascading errors: To estimate the error propagation of misclassification at one stage, we quantified the effect of misclassification on downstream predictions, and there was little error propagation, which quantifies the efficacy of hierarchical loss -balancing. Text sequences are shortened into 128 tokens with semantic completeness and an increase in memory efficiency. GrievEase combines image embeddings based on CLIP with engineered features into an XGBoost classifier on multimodal validation to ensure that uploaded evidence is consistent with complaint text. The grid search using learning-rate warmup, dropout, and gradient clipping was used to fine-tune the hyperparameters to train it stably. SeamlessM4T facilitates multilingual voice support. Having Fastapi, PostgreSQL, and a ReactJS dashboard, the platform provides a transparent, efficient, and scalable grievances redressal system that builds a hybrid approach of textual and visual intelligence to ensure the provision of quick, efficient judiciary.

Fields of Study & Subjects

Text and Document Classification Technologies Artificial Intelligence in Law Topic Modeling

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
01–08
Publication date
21 January 2026
Language
EN