Emotional Plan Supporter (EPS)

📘 Overview

EPS is an AI-powered assistant designed to assist individuals facing challenges in long-term planning. It combines emotional support with actionable strategies to offer practical solutions and motivation.

💡 Motivation

This project was developed during the first year, second semester of a Master's program in Artificial Intelligence as part of a large language model term project. Inspired by personal challenges with ADHD, I combined my interest in Emotional Support Conversations (ESC) with practical planning tools to address the complexities of long-term goal-setting.

🧩 Components

The project consists of two main components:

1. LLMTermProjEPS: A module focused on empathetic planning support using OpenAI GPT.
2. BlenderBot: A conversational chatbot module based on Facebook’s Blenderbot.

---

🔧 Project Structure

LLMTermProjEPS/
├── EPS_MVP.py          # Empathetic Support Planner script
├── requirements.txt    # Python dependencies for this module

BlenderBot/
├── Blenderbot_model.py # Blenderbot-based chatbot implementation
├── environment.yml     # Conda environment configuration
├── version_check.py    # Python version and library check script

---

🚀 Installation and Setup

1. Clone the Repository

git clone https://github.com/Yerin99/Emotional-Plan-Supporter.git
cd Emotional-Plan-Supporter

2. Setting up the Environment

For BlenderBot:

1. Create a Conda environment using the provided environment.yml:

   conda env create -f BlenderBot/environment.yml
   conda activate blenderbot_env

For LLMTermProjEPS:

1. Install dependencies from requirements.txt:

   pip install -r LLMTermProjEPS/requirements.txt

---

▶️ How to Run

1. Run the Empathetic Support Planner (EPS_MVP):

python LLMTermProjEPS/EPS_MVP.py

2. Run the BlenderBot Chatbot:

python BlenderBot/Blenderbot_model.py

---

🌟 Features

• EPS_MVP:
  • Provides detailed, empathetic, and actionable plans based on user input.
  • Supports SMART (Specific, Measurable, Achievable, Relevant, Time-bound) framework for task planning.
• BlenderBot:
  • Engages in natural and interactive conversations using the Blenderbot model.
  • Leverages Huggingface Transformers library and PyTorch.
  • Note: BlenderBot may not handle long instructions well. Keep inputs concise for optimal results.

---

⚠️ Limitations

• BlenderBot:
  • While effective for empathetic conversational support, BlenderBot was found unsuitable for planning tasks due to its difficulty in understanding and processing lengthy prompts effectively.

---

📋 Requirements

LLMTermProjEPS

• Python 3.8+
• See LLMTermProjEPS/requirements.txt for details.

BlenderBot

• Python 3.8+
• Conda environment (environment.yml) includes:
  • PyTorch
  • Transformers
  • Other required dependencies

---

📄 Documentation

For a detailed explanation of the project, including the following:

1. Project topic and problem statement
2. Challenges and methods to address them
3. Experimental results and analysis
4. Conclusion, limitations, and future directions
5. References

Please refer to the Final Report: Emotional Plan Supporter.

The report also includes appendices with real usage examples and illustrative images, along with details on the SMART framework.

---

📜 License

This project is licensed under the MIT License due to the usage of Facebook’s Blenderbot, which is distributed under an MIT license. See LICENSE for details.

---

🙏 Acknowledgments

• OpenAI for GPT API
• Facebook AI for Blenderbot
• Huggingface for the Transformers library

---