Using LLMs in huggingface for sentiment analysis, translation, summarization and extractive question answering

AI-powered citizen support platform built with Flask. Modules include civic issue reporting, legal Q&A AI, smart budgeting AI, health prediction, and more.

The Recursive Semantic Refinement Network is an approach to abstractive summarization that frames the task as an iterative regression problem in a semantic space. Instead of generating text word-by-word, the network refines a fixed-size semantic embedding of the summary until it converges to a state semantically close to the ground-truth embedding.

This project is a desktop-based application that allows users to simulate a chat environment and summarize the entire conversation using modern Natural Language Processing techniques. It is designed to help users quickly grasp the essence of long chat threads without having to read every message.