Discord ChatBot with Natural Language Processing and Crypto Data Features

This project creates a chatbot for Discord with multiple advanced features, integrating real-time cryptocurrency data and various Natural Language Processing (NLP) capabilities.

Features

1. Crypto Information Retrieval:

   • !run command to provide users with real-time and historical cryptocurrency data.

2. Web Crawling and Search:

   • Implementation of web crawling capabilities to build a comprehensive database for information retrieval.

3. Content Filtering:

   • Development of a filtering mechanism to exclude negative content from the downloaded web pages using a trained Neural Network negativity classifier.

4. Content Generation:

   • Implementation of a neural network trained on the crawled data to generate content (GenAI).

5. Advanced Content Generation:

   • Enhanced content generation utilizing GPT-2 via HuggingFace transformers.

DevLogs - Short texts explaining the development process for each of the feature iterations of the chatbot.

• DevLog 0 - Starting the bot structure!
• DevLog 1 - Adding the !run command to send crypto information to the user.
• DevLog 2 - Adding web crawling and search functions to the formed database.
• DevLog 3 - Filter by negativity on the content of downloaded pages.
• DevLog 4 - Content generation from the crawled database.
• DevLog 5 - Content generation with GPT (Gpt-2 using HuggingFace transformers)

How to Run Locally

1. Clone the repository:

   git clone https://github.com/RicardoRibeiroRodrigues/NLP-DiscordBot

2. Install the dependencies:

   pip install -r requirements.txt

3. Add the bot's token and the CoinCap API token to a .env file in the following format:

   export TOKEN=bot_token_here
   export API_KEY=api_key_here

4. Run the setup script (this script will download WordNet):

   python3 manager_script.py setup

5. Run the bot:

   python3 main.py

   PS: The file `manager_script.py` has a cleanup function to clean all generated data in `data/` and `models/`.

Author

• Ricardo Ribeiro Rodrigues - [email protected]