📄 Sentiment Analysis with BERT & XAI Project

---

📌 Description

This project includes a Jupyter Notebook that performs sentiment analysis using the BERT model, while also leveraging XAI (Explainable AI) techniques to make the model's results more transparent and interpretable.

🔹 Key Features:

• Use of BERT for text sentiment analysis
• Application of XAI techniques such as LIME and SHAP for model interpretability
• Visualization of word impact on model predictions
• Ability to train and evaluate the model on custom datasets

---

🔧 Installation & Requirements

Dependencies

Libraries used in this project:

• transformers
• torch
• numpy and pandas
• lime and shap
• matplotlib and seaborn

If any library is not installed on your system, you can install it using:

pip install transformers torch numpy pandas lime shap matplotlib seaborn

Run the Notebook

jupyter notebook sentiment-analysis-bert-xai.ipynb

---

📊 How It Works

1️⃣ Load Data 📥

• Text data for sentiment analysis is loaded.

2️⃣ Preprocess Data 🔄

• Data is cleaned and prepared for the model.

3️⃣ Predict Sentiment with BERT 🤖

• Input text is processed by the model, and its sentiment is identified.

4️⃣ Model Interpretability Analysis (XAI) 🔍

• The impact of each word on the prediction is analyzed using XAI techniques like LIME and SHAP.

---

📌 Sample Output

✅ Sentiment Prediction:

Text	BERT Prediction
"I love this product!"	Positive
"The service was terrible."	Negative

✅ Model Interpretability (LIME & SHAP):
📊 SHAP and LIME visualizations show which words had the most influence on the prediction.

---

• If you have suggestions for improving the project, please submit a Pull Request.
• To report issues, please open an Issue.