EU-Chatbot-Reco-System
π The Challenge:
Compliance with EU Commission environmental carbon emission reporting mandates can be a laborious process, prone to human error, and costly in terms of time and resources.
π€ The Solution:
Introducing the WhatsApp-based chatbot designed to streamline and simplify carbon emission reporting, ensuring compliance with EU Commission standards.
DEMO:
GPT DEMOS
Main Development Branch for each model:
GPT - gpt-bot
Llama2 - llama2-bot
Custom Mistral 7B - custom-bot
π Key Features & Benefits:
- Dynamic Reporting: Utilise Natural Language Processing (NLP) and AI to send and receive text messages via WhatsApp, which auto-populate the carbon emission report with real-time data.
- Efficiency: Eliminate manual data entry and reduce reporting time significantly.
- User-Friendly: No complex installations or training required; WhatsApp familiarity ensures ease of use.
- Accuracy & Compliance: Ensure precise reporting to meet EU Commission requirements, minimising the risk of penalties.
- Real-Time Updates: Receive notifications and alerts regarding reporting deadlines and regulatory changes.
- Cost-Effective: A cost-efficient alternative to hiring additional personnel or outsourcing reporting.
- Environmental Impact: Simplifying reporting contributes to a company's broader sustainability efforts.
π€ Affected Stakeholders:
- Organizations and Businesses: Companies subject to EU carbon emission reporting mandates benefit from efficient and accurate reporting, reducing compliance risks.
- Regulatory Authorities: EU Commission and national environmental agencies benefit from improved data collection and compliance verification.
- Environmental Advocacy Groups: Organizations focused on sustainability appreciate efforts to reduce carbon emissions and improve reporting accuracy.
- Investors and Shareholders: Increased reporting accuracy enhances investor confidence and supports sustainability goals.
- Environmental Consultants and Auditors: These experts see value in a tool that streamlines reporting processes and enhances accuracy.
π οΈ Technology Stack:
- Programming Languages: Python
- Frameworks: Flask
- Natural Language Processing Libraries: NLTK, Huggingface, Langchain
- Database: PostgreSQL
- Messaging Platform: WhatsApp API
- Machine Learning:
- Data Visualization: Matplotlib
π§ Methodology:
Our approach involves:
- Data collection from various sources, including public agricultural ministry datasets and internal databases.
- Training and fine-tuning NLP models for text analysis.
- Developing a chatbot using Python and Flask, including defining user flow and minimizing questions asked until report is filled.
- Integration with WhatsApp API for user interaction.
π Data Sources:
We collect data from:
- Public agricultural ministry datasets
- EU Commission guidelines and datasets.
- Internal databases from user intraction
Fake Template Report: https://drive.google.com/file/d/1YsBcwr6d59_d4rpqOVz3nuBYTawpmfz9/view?usp=sharing
π Implementation Details:
The chatbot is implemented using Flask and integrated with the WhatsApp API to enable real-time reporting. NLP models are used to analyze text messages and auto-populate reports.
π§ͺ Testing and Validation:
The chatbot was rigorously tested against various reporting scenarios, and validation was performed to ensure it meets EU Commission standards for accuracy and compliance.
π Results and Metrics:
The EcoReportBot has reduced reporting time and improved accuracy to ensure compliance with EU Commission mandates. Key metrics include response time and user satisfaction.
π‘ Lessons Learned:
π Future Enhancements:
In the future, we plan to:
- Expand language support.
- Incorporate advanced AI features for predictive analysis.
- Develop a mobile app for accessibility.
- Enhance reporting visualization.
π Documentation:
User guides and technical documentation can be found here.
π₯ The Team:
- Adnan Bhanji: Project Manager
- Beatrice Mossberg: Data Engineer
- Riyad Mazari: Data Scientist
- Sofia Morena Lasa: Data Scientist
- Khaled Akel: Machine Learning Engineer
- Hussein Soliman: MLOps Engineer
π Acknowledgments:
...
We would like to express our gratitude to our mentors, collaborators, and data providers who contributed to the success of this project.
π Appendix:
For additional code snippets, data samples, and graphs, please refer to the appendix.