https://github.com/ninimecri/Natural-Language-Processing-NLP-sprint 15 project.ipynb

Natural-Language-Processing-NLP-
This project aims to build a sentiment classification system for movie reviews using multiple NLP techniques and models. It was developed as part of a challenge for Film Junky Union, a community for classic film enthusiasts. The objective is to classify IMDB reviews as positive or negative, achieving a minimum F1-score of 0.85.

⚙️ Key Features
Five different models ranging from a simple baseline to deep learning with BERT.
A single unified function that allows easy switching between models for training and testing.
Ability to classify new custom reviews using all trained models.
Modular and reproducible code with clearly separated sections for preprocessing, training, and evaluation.

🧠 Models Implemented
Model Description
Model 0 Dummy Classifier (Baseline)
Model 1 NLTK preprocessing + TF-IDF + Logistic Regression
Model 2 spaCy preprocessing + TF-IDF + Logistic Regression
Model 3 spaCy preprocessing + TF-IDF + LightGBM Classifier
Model 4 BERT Transformer (fine-tuned for sentiment classification)
All models are run using a custom function, which encapsulates the full pipeline: preprocessing → vectorization → training → evaluation.

📥 Dataset
The dataset used consists of movie reviews from IMDB, with labeled sentiment polarity (positive/negative). It is loaded and split into training and test sets during execution.

📈 Evaluation Models are evaluated using standard metrics:

• Accuracy
• Precision
• Recall
• F1-score (target ≥ 0.85)

💡 Future Improvements Hyperparameter optimization (especially for LightGBM and BERT)

Incorporation of more robust datasets

API or web-based deployment using Streamlit or Flask

📚 Dependencies Python 3.8+

scikit-learn

NLTK

spaCy

LightGBM

Transformers (Hugging Face)

pandas, NumPy, matplotlib