StockPulse is an intelligent stock sentiment analysis tool that extracts and analyzes financial sentiment from online articles using fine-tuned language models. It helps traders, researchers, and analysts understand how specific stock tickers are discussed in news content.
-
🔍 URL-based Sentiment Extraction
Input a news article URL — StockPulse scrapes and processes the content automatically. -
🧠 NER-powered Ticker Matching
Automatically detects relevant company names and maps them to their stock tickers. -
💬 Fine-Tuned Sentiment Classification
Uses a BERT-based model to classify sentiment as Positive, Negative, or Neutral for each detected ticker. -
📈 Confidence Score Output
Each sentiment prediction is paired with its associated confidence level.
- Python
- Hugging Face Transformers
- Torch / PyTorch
- Newspaper (for web scraping)
- Streamlit (for interactive demo)
- User provides a URL
- The article is scraped and cleaned
- Entities (company names) are extracted
- Mapped to stock tickers
- Sentiment analysis is performed per ticker
- Returns: Ticker, Sentiment, Confidence
1. Ticker: AAPL
Sentiment: Positive
Confidence: 87.32%
2. Ticker: TSLA
Sentiment: Negative
Confidence: 91.44%🔗 Streamlit Demo: https://stockpluse.streamlit.app/
Try the interactive demo to explore StockPulse in action!
- Pretrained BERT models from Hugging Face 🤗
- Yahoo Finance for ticker data reference
- FinBERT and Financial Sentiment resources
- ✅ Improved NER mapping
- Add multi-URL batch processing
- Deploy on cloud with live news integration