10-k Filings Analysis

SigmaNLP is a Streamlit-based application designed for performing sentiment analysis on 10-K filings of publicly traded companies. Using advanced natural language processing (NLP) techniques, including FinBERT, SigmaNLP analyzes various sections of 10-K filings to compute sentiment ratings and visualize trends over time. Additionally, it fetches related news articles using NewsAPI to provide additional context for the analyzed company.

Features

• Sentiment Analysis:
  • Analyze specific sections of 10-K filings (e.g., "Business Overview", "Risk Factors").
  • Uses FinBERT, a pre-trained sentiment analysis model for financial text.
• Visualization:
  • Displays overall sentiment trends over the years using a line graph.
  • Provides a color-coded table for easy interpretation of sentiment scores.
• News Integration:
  • Fetches and displays related news articles for the company ticker using NewsAPI.

Installation

To run SigmaNLP locally, follow these steps:

Prerequisites

1. Python 3.8 or higher.
2. Git.
3. Streamlit.
4. NewsAPI Key (required for fetching related news articles).
5. TextRazor API Key (optional, for extended text analysis).

Clone the Repository

git clone https://github.com/varchan123/sigmanlp.git
cd sigmanlp

Set Up a Virtual Environment

It is recommended to use a virtual environment for better dependency management:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

Install Dependencies

Install the required Python dependencies from the requirements.txt file:

pip install -r requirements.txt

Export API Keys

Ensure you export your NewsAPI and TextRazor API keys as environment variables:

export NEWSAPI_KEY=your_newsapi_key
export TEXTRAZOR_KEY=your_textrazor_key

Replace your_newsapi_key and your_textrazor_key with your actual API keys.

Usage

To run the application locally, execute the following command:

streamlit run streamlit_app.py

Once the application is running, visit http://localhost:8501 in your web browser to access the interface.

How It Works

1. Input:

   • Enter the stock ticker (e.g., GOOG for Google) and the start year (e.g., 2019) in the input fields.

2. Processing:

   • The application downloads 10-K filings using an Edgar crawler.
   • Sentiment analysis is performed on specific sections of the filings using FinBERT.

3. Output:

   • A line graph shows the overall sentiment trend over the years.
   • A table displays sentiment scores for each section, color-coded for better clarity.
   • Related news articles are listed at the bottom of the interface.

File Structure

• streamlit_app.py: Main Streamlit application code.
• requirements.txt: Python dependencies.
• edgar-crawler/: Contains scripts for downloading and processing 10-K filings.
• datasets/: Stores raw and extracted filings data.

Example Workflow

1. Enter the stock ticker (e.g., AAPL) and the start year (e.g., 2020).
2. Click the Analyze button.
3. The application will:
   • Download and process the 10-K filings.
   • Perform sentiment analysis.
   • Display the results as a graph and table.
   • Fetch related news articles for the specified company.

Key Technologies

1. Streamlit: Provides the interactive web-based user interface.
2. FinBERT: A transformer-based model fine-tuned for sentiment analysis on financial text.
3. NewsAPI: Fetches related news articles to provide additional insights.
4. TextRazor (Optional): Advanced text analysis API for extracting topics, entities, and summaries.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests for bug fixes or new features.

Support

For issues or questions, please open an issue in the GitHub repository.