Skip to content

Repository files navigation

NewsNER

Overview

The NewsNER project is a groundbreaking initiative that leverages the power of transformer-based models, fine-tuned specifically for an extensive Hindi news dataset. This project embodies the cutting-edge in Natural Language Processing (NLP), providing comprehensive analytics and classification of news content.

Key Features

Transformer-Based Models

Our core technology is based on advanced transformer models, renowned for their effectiveness in understanding and generating human language. These models have been meticulously fine-tuned on a diverse range of Hindi news data, ensuring high accuracy and relevance in the context of Indian news media.

Streamlit Frontend

  • User-Friendly Interface: The project employs Streamlit for its frontend, providing an intuitive and interactive user interface that simplifies the interaction with our complex models.
  • Seamless Integration: This interface seamlessly connects with the backend, allowing users to effortlessly send requests and receive insights.

Backend Capabilities

The robust backend of NewsNER is designed to handle a variety of NLP tasks:

  1. Live POS (Part-of-Speech) Tagging:

    • Dynamically identifies grammatical groups of words in Hindi news text.
    • Enhances understanding of linguistic structures in news articles.
  2. NER (Named Entity Recognition) Tagging:

    • Efficiently recognizes and categorizes key entities in the news text such as names, organizations, locations, etc.
    • Critical for detailed news analysis and information extraction.
    • Done using HMM and CRF Method (python-crfsuite)
  3. Zero-Shot Text Classification:

    • Classifies news into various categories without needing labeled examples for every class.
    • Empowers the model to handle a wide range of topics and themes.
  4. Text Summarization:

    • Generates concise summaries of news articles.
    • Allows quick understanding of news content without reading the full text.

Dataset

The backbone of NewsNER is a comprehensive Hindi news dataset, encompassing a wide array of topics and writing styles. This rich dataset ensures that our models are well-adapted to the nuances of Hindi language and the Indian news landscape.

Applications

NewsNER is ideal for journalists, researchers, and media monitoring agencies, offering deep insights into news trends, narrative analysis, and content summarization. It's also an invaluable tool for anyone interested in understanding the complex landscape of Hindi news media.


For more information or to request a demo, please visit our website or contact us at pranav_deepak@rocketmail.com.

Releases

Packages

Contributors

Languages