Sentiment Analysis of Romanized Kannada Product Reviews

Flask Web Application

JSS Science & Technology University, Mysuru
Dept. of Computer Science & Engineering — Final Year Project 2024–25

---

Folder Structure

sentiment_analysis_flask/
│
├── app.py                        ← Flask entry point  →  python app.py
│
├── templates/
│   ├── base.html                 ← Shared sidebar layout
│   ├── index.html                ← Homepage / portfolio
│   ├── analyzer.html             ← Single review prediction
│   ├── bulk.html                 ← CSV/Excel bulk analysis
│   └── history.html              ← Prediction history dashboard
│
├── static/
│   ├── css/style.css             ← Full stylesheet
│   └── js/main.js                ← Shared JS utilities
│
├── utils/
│   ├── preprocessing.py          ← Text cleaning pipeline
│   ├── model_loader.py           ← SVM / BiLSTM / mBERT loaders
│   ├── keyword_rules.py          ← Keyword pre-prediction engine
│   └── database.py               ← SQLite history store
│
├── models/                       ← Place your trained model files here
│   ├── svm_model.pkl
│   ├── tfidf_vectorizer.pkl
│   ├── bilstm_model.h5
│   ├── bilstm_tokenizer.pkl
│   ├── bilstm_label_encoder (1).pkl
│   └── mbert_model.pkl
│
├── dictionary.csv                ← Normalization dictionary
├── history.db                    ← Auto-created on first run
└── requirements.txt

---

Setup Instructions

Step 1 — Create virtual environment

cd C:\Users\sindhu\Desktop\sentiment_analysis_flask

python -m venv venv

# Activate on Windows:
venv\Scripts\activate

# Activate on Mac/Linux:
source venv/bin/activate

Step 2 — Install dependencies

pip install -r requirements.txt

If you already have a working myenv from the Streamlit version, you only need to add Flask:

pip install flask

Step 3 — Add your model files

Copy these files into the models/ folder:

File	Description
svm_model.pkl	Trained SVM classifier
tfidf_vectorizer.pkl	TF-IDF vectorizer
bilstm_model.h5	BiLSTM Keras model
bilstm_tokenizer.pkl	BiLSTM tokenizer
bilstm_label_encoder (1).pkl	Label encoder
mbert_model.pkl	Fine-tuned mBERT

Also place dictionary.csv in the root folder.

Step 4 — Run the application

python app.py

Open your browser at: http://localhost:5000

---

Pages

URL	Page	Description
/	Home	Academic portfolio — abstract, dataset, models, team
/analyzer	Sentiment Analyzer	Single review prediction with confidence charts
/bulk	Bulk Analysis	Upload CSV/Excel, predict all rows, download results
/history	Prediction History	Browse, filter, export all past predictions

---

API Endpoints

Method	Endpoint	Description
POST	/api/predict	Single review prediction
POST	/api/bulk	Bulk file analysis
GET	/api/history	Get all history records
DELETE	/api/history/delete/<id>	Delete one record
POST	/api/history/clear	Clear all records
GET	/api/history/export	Download history CSV

Example API call (predict):

import requests

response = requests.post("http://localhost:5000/api/predict", json={
    "text":  "product tumba chennagide delivery fast bantu",
    "model": "SVM"   # or "BiLSTM" or "mBERT"
})

print(response.json())
# {
#   "label": "Positive",
#   "confidence": 92.3,
#   "proba": {"Negative": 2.1, "Neutral": 5.6, "Positive": 92.3},
#   "clean_text": "product tumba chennagide delivery fast bantu",
#   "tokens": ["product", "tumba", "chennagide", "delivery", "fast"],
#   "model": "SVM",
#   "method": "keyword"
# }

---

Difference from Streamlit Version

Feature	Streamlit	Flask
Frontend	Python widgets	HTML + CSS + JS
Styling	Streamlit CSS	Custom Inter CSS
Charts	Plotly	Chart.js
API	None	REST JSON API
Deployment	streamlit run	python app.py
Port	8501	5000

All backend logic (preprocessing, models, database) is identical.

---

Important Notes

1. sklearn version — Models were saved with sklearn 1.6.1. Run:

   pip install scikit-learn==1.6.1

2. mBERT on CPU — The app forces CPU loading automatically. No GPU required. First load takes ~30 seconds.

3. mBERT tokenizer — Downloaded from HuggingFace on first run (~500KB, cached locally after that).

4. history.db — Created automatically in the project root on first prediction.

---

Run in Production (optional)

pip install gunicorn

gunicorn -w 2 -b 0.0.0.0:5000 app:app

---

JSS S&T University, Mysuru — CS & Engineering — 2024–25