chennab28/movie-review-sentiment

🎬 Movie Review Sentiment (Mini)

A tiny hand-built dataset of one-sentence movie reviews labeled with their sentiment, used to demonstrate the full Hugging Face workflow:

Dataset → Model → Space

What's inside

Each example has two fields:

• —text — an English sentence reviewing a movie
• —label — one of positive, negative, neutral

Why this dataset exists

This dataset is intentionally small and clean so a tiny model (bert-tiny, ~2 encoder layers) can fine-tune on it quickly, even on a CPU. It powers the movie-review-sentiment classifier and its companion Gradio Space.

How it was built

data/build_dataset.py combines:

1. 1.Hand-written seed reviews for each class (realistic sentences),
2. 2.Template-generated variations that swap adjectives / phrases to give the model more signal,
3. 3.A deterministic 80/20 train/test split (random.seed(42)).

You can rebuild it yourself:

python data/build_dataset.py

Load it

from datasets import load_dataset

ds = load_dataset("chennab28/movie-review-sentiment")
print(ds["train"][0])   # {'text': ..., 'label': 'positive'}

Feedback

Try the live demo in the accompanying Space or improve the dataset and open a pull request!