🚀 Introducing Meowsenti 🚀

Meowsenti is a simple application designed to analyze sentiment in Indonesian mobile banking applications.

🤖 Model Overview

This system is built using two popular Natural Language Processing algorithms: Multinomial Naïve Bayes and Support Vector Machine (SVM). The model has been trained on a dataset of 13,527 reviews from the BRImo mobile banking application.

The dataset is split into 60% for training and 40% for testing.

🔍 MFSvc Model (Support Vector Machine)

The MFSvc model is optimized using GridSearchCV. Below is the parameter grid used:

Parameter	Values
C	0.1, 1, 10
kernel	linear, rbf
gamma	scale, auto
class_weight	balanced, None

After hyperparameter tuning, the best configuration is:

Best parameters found: {'C': 10, 'class_weight': 'balanced', 'gamma': 'scale', 'kernel': 'rbf'}
Best cross-validation accuracy: 95.67%

📊 Evaluation Metrics

Class	Precision	Recall	F1-score
Positive (1)	0.9226	0.9179	0.9202
Negative (0)	0.6281	0.6281	0.6281
Neutral (2)	0.7348	0.7436	0.7392

• Train Accuracy: 95.67%
• Test Accuracy: 83.11%

🔥 Visualizations

• Classification Report:

  

• Confusion Matrix:

  

📉 MFNb Model (Naïve Bayes)

The MFNb model is also optimized using GridSearchCV. Below is the parameter grid used:

Parameter	Values
alpha	0.01, 0.1, 1, 10
fit_prior	True, False

After hyperparameter tuning, the best configuration is:

Best parameters found: {'alpha': 0.01, 'fit_prior': True}
Best cross-validation accuracy: 95.23%

📊 Evaluation Metrics

Class	Precision	Recall	F1-score
Positive (1)	0.9012	0.9338	0.9172
Negative (0)	0.6507	0.6006	0.6246
Neutral (2)	0.7386	0.7085	0.7232

• Train Accuracy: 95.23%
• Test Accuracy: 82.77%

🔥 Visualizations

• Classification Report:

  

• Confusion Matrix:

  

---

🛠️ How to Run Meowsenti Locally

🐳 Using Docker

First, pull the image and run:

# Pull the Docker image
docker pull mframadann/meowsenti

# Run the container
docker run -d --name sentiment-app -p 5001:7860 mframadann/meowsenti

🔬 Testing the API

Use tools like Postman or Insomnia, or send a request via cURL:

curl -X POST \
  http://localhost:5001/api/v1/analyze-sentiment \
  -H "Content-Type: application/json" \
  -d '{
        "reviews": [
           {"review": "Bagus banget aplikasinya"},
           {"review": "Aplikasi apa ini, jelek banget idih"},
           {"review": "Cukup membantu saya bertransaksi"},
           {"review": "Mantap banget euy aplikasinya"}
        ],
        "model": "MFNb"
    }'

📡 API Response

{
  "data": {
    "sentiment": [
      {
        "alg_type": "Naïve Bayes",
        "kind_of_sentiment": "Positive",
        "review": "Bagus banget aplikasinya"
      },
      {
        "alg_type": "Naïve Bayes",
        "kind_of_sentiment": "Negative",
        "review": "Aplikasi apa ini, jelek banget idih"
      },
      {
        "alg_type": "Naïve Bayes",
        "kind_of_sentiment": "Neutral",
        "review": "Cukup membantu saya bertransaksi"
      },
      {
        "alg_type": "Naïve Bayes",
        "kind_of_sentiment": "Positive",
        "review": "Mantap banget euy aplikasinya"
      }
    ]
  },
  "status": "success"
}

💻 Running Meowsenti Manually

Clone the Repository

git clone https://github.com/mframadann/meowsenti.git && cd meowsenti

Set Up a Virtual Environment

python3 -m venv .venv
source .venv/bin/activate

Install Dependencies

pip install --no-cache-dir -r requirements.txt

Start the Application

python app.py ## run on http://localhost:7860

---

📜 API Documentation

For detailed API documentation, visit: Mframadan Labs. You can also try Meowsenti directly on Hugging Face Spaces:

🔗 Meowsenti on Hugging Face

⚠️ Note

This model works only for Indonesian text and last updated of analytics report is based on model observation in oct 2024. Thanks for visiting! Feel free to connect with me:

🔗 LinkedIn

---