SpamBot - Your Daily Activity Summarizer 📋✨

Live Demo 🚀: SpamBot

Tools and Technologies 🛠️

• Language: Python
• Libraries: Streamlit, Transformers, Langchain, ChromaDB and Google AI Studio
• Models: gemini-pro, bart-large-cnn and distilbert-base-cased-distilled-squad

Background

SpamBot is an integral part of a VR-based project designed to help users summarize their daily activities and perform QA tasks. It aims to enhance productivity and assist in recalling minutes of meetings. Initially, we developed a VR-based multilingual speech-to-text transcriber and intended to integrate SpamBot with it. However, we encountered challenges in connecting these technologies.

What is SpamBot? 🤖

SpamBot is a standalone application that takes a transcript.txt file as input, allowing users to perform QA and summarization tasks. This makes it easier to summarize lengthy conversations and recall important meeting details.

System Architecture 🏗

Fig 1: Basic RAG pipeline using Langchain.

Description 📜

• qachat.py: Contains foundational transformers like facebook/bart-large-cnn for summarization and distilbert-base-cased-distilled-squad for QA. Due to the foundational nature of these models, their results were not very accurate. Hence, we upgraded to using gemini-pro for QA and t5-base for summarization.
• main.py: The core file of the application, utilizing gemini-pro for both QA and summarization tasks. Users can input a transcript.txt file to perform these tasks.
• requirements.txt: Lists all the necessary libraries to run the application.

How to Run? 🚀

1. Clone the repository.
2. Install the required libraries using the command: pip install -r requirements.txt
3. Save your API key in a .env file.
4. Run the main.py file on Streamlit using the command: streamlit run main.py

Note:

• You can modify the transcript.txt file to test the application with different inputs.

• Feel free to test the qachat.py file with different models and transformers to explore more functionalities.

Future Work 🛠️

• Integrate SpamBot with the VR-based transcriber.
• Enhance the summarization and QA models for improved accuracy.
• Add more functionalities like speech-to-text conversion and text-to-speech synthesis.

Let's make your daily activities more manageable with SpamBot! 🌟

Created by Team SpamBots with ❤️:

• Janesh Kapoor (Team Lead)
• Divyam Khorwal
• Tushar Chandra
• Varsha Bhaskar