diff --git a/README.md b/README.md
index 868c05a..0ef3bc4 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,45 @@
-# sentiment_product
\ No newline at end of file
+Certainly! Below is a template for a structured and informative README file for your sentiment analysis project. I've included sections on project overview, dataset, model fine-tuning, running evaluations, and additional notes. Please customize it further based on your specific project details.
+
+---
+
+# Sentiment Analysis of Customer Reviews
+
+## Overview
+
+This project focuses on sentiment analysis of customer reviews using machine learning techniques. The goal is to fine-tune various models on a customer review dataset, compare their performance, and evaluate pre-built models for sentiment analysis. The project is designed to be executed on CPU, utilizing low-size models for efficiency.
+
+## Dataset
+
+The dataset used for this project consists of approximately 20,000 customer comments. The labels assigned to each comment range from -0.9 to 0.9, reflecting a sentiment score. The sentiment scores are mapped to a classification scale resembling a star rating, ranging from 1 star to 5 stars.
+
+## Model Fine-Tuning
+
+### Fine-Tuned Model
+
+The project employs the 'distilbert-base-uncased' model from Hugging Face as the base model for fine-tuning. This model is chosen for its lower size, making it suitable for training and testing on CPU.
+
+To fine-tune the model, the sentiment labels are transformed into a classification task, converting the sentiment scores into star ratings. The fine-tuning process is implemented in the `evaluation.py` script, providing insights into the model's performance.
+
+### Pre-Built Model
+
+Additionally, a pre-built classification model, 'nlptown/bert-base-multilingual-uncased-sentiment,' is included in the evaluation process. This model is specifically fine-tuned for customer reviews.
+
+## Running Evaluations
+
+To evaluate the fine-tuned and pre-built models, run the `evaluation.py` script. This script generates confusion matrices for each model, facilitating a comparative analysis of their performance. The confusion matrices offer a visual representation of how well the models classify sentiments across different classes.
+
+```bash
+python evaluation.py
+```
+
+## Additional Notes
+
+- Ensure that the required dependencies are installed. Refer to the `requirements.txt` file for details.
+- The project is optimized for execution on CPU and utilizes smaller-sized models for efficiency.
+- Customize hyperparameters, model architecture, or training settings as needed for your specific requirements.
+
+Feel free to explore and experiment with different models and datasets to enhance the project's performance.
+
+---
+
+You can copy and paste this template into your README file and further modify it based on any additional information you'd like to include or specific details about your project.