News Summarization & Hindi TTS Application

Overview

This project extracts news articles related to a company, summarizes them, performs sentiment analysis, and generates a Hindi audio report.

Key Features:
Web Scraping for latest news (BeautifulSoup)
Summarization using Transformers (Hugging Face)
Sentiment Analysis (Positive, Negative, Neutral)
Comparative Sentiment Insights
Hindi TTS Audio Output (gTTS)
API Support for analysis & querying API Support using Flask (api.py)
Web UI using Gradio (app.py)

Project Setup, Model Details, API Development, API Usage, Assumptions & Deployment

Project Setup

Clone the Repository

git clone https://github.com/guruprashanth2004/news-summarization-tts.git
cd news-summarization-tts

set up virtual environment

python -m venv venv
venv\Scripts\activate

Install Dependencies

pip install -r requirements.txt

Run the API

python api.py

Run the Application

python app.py

Model Details

Summarization Model

• Purpose: Summarizes news articles into concise summaries.
• Library Used: transformers

Sentiment Analysis Model

• Purpose: Analyzes sentiment of news articles.
• Library Used: transformers

Text-to-Speech (TTS) Model

• Model: gTTS (Google Text-to-Speech)
• Purpose: Converts summarized sentiment analysis into Hindi speech.
• Library Used: gtts
• Example Input: "Tesla की हालिया खबरें ज्यादातर सकारात्मक हैं।"
• Example Output: output.mp3 (Hindi speech file)

---

**API Development (api.py)

The backend API is built using Flask and provides two endpoints:

Analyze News (/analyze)

Purpose: Fetch news, summarize, analyze sentiment, and generate Hindi TTS.
Method: GET
URL:
http://127.0.0.1:5000/analyze?company=Tesla Response:

{
    "articles": [
        {
            "title": "Tesla Stock Hits New Highs",
            "summary": "Tesla's stock price reached record levels...",
            "sentiment": "Positive"
        }
    ],
    "comparison": {
        "Sentiment Distribution": {"Positive": 1, "Negative": 0, "Neutral": 0}
    },
    "audio_file": "output.mp3"
}

Query News by Sentiment (/query)

Purpose: Fetch articles based on sentiment filter. Method: GET URL: http://127.0.0.1:5000/query?company=Tesla&sentiment=Positive Response:

{
    "articles": [
        {
            "title": "Tesla's Profits Rise",
            "summary": "Tesla reports record-breaking profits...",
            "sentiment": "Positive"
        }
    ]
}

Third-Party APIs & Tools Used

API/Tool	Purpose
Bing News Scraping	Extracts latest news articles
Hugging Face Transformers	Summarization & Sentiment Analysis
Google Text-to-Speech (gTTS)	Converts sentiment report to Hindi speech

Assumptions & Limitations

Assumptions

• The news articles contain relevant information about the company.
• Sentiment analysis is reasonably accurate, but may misinterpret sarcasm.
• The Hindi TTS model correctly pronounces the text, but may have intonation issues.

Limitations

• Web Scraping Issues:
• Some websites block bots (may require rotating proxies).
• Bing News doesn’t always provide full article content.
• Sentiment Model Limitations:
• Works well for English news, but doesn’t analyze Hindi news.

• works better when deployed in vs code than in hugging face

Deployment

This application is deployed on Hugging Face Spaces.

🔗 Live Demo: https://huggingface.co/spaces/guru14102004/news-summarization-tts