Udacity Deep Learning Nanodegree (nd101) — Course 4: RNNs and Transformers, final project.

Analyze sentiment of movie reviews in three languages (English, French, Spanish):

1. Translation — fr→en and es→en via HuggingFace Helsinki-NLP/opus-mt-* MarianMT models
2. Merge — combine the three CSVs into one dataframe
3. Sentiment — binary classification via DistilBERT (distilbert-base-uncased-finetuned-sst-2-english)
4. Export — result/reviews_with_sentiment.csv

30 reviews (10 per language) → 15 Positive / 15 Negative.

• Python 3.14 / PyTorch 2.10 / HuggingFace transformers 5.3
• Runs on CPU or MPS (Apple Silicon); no GPU required

pip install transformers sentencepiece sacremoses pandas torch
python run_project.py   # inference-only, ~2-3 min

• project.ipynb / project.html — executed submission notebook
• run_project.py — standalone training/inference script
• data/movie_reviews_{eng,fr,sp}.csv — source reviews
• result/reviews_with_sentiment.csv — final output

Part of Leo's Udacity Masters in AI journey. See also:

• landmark-classifier — Course 3 CNN project
• face-generation — Course 5 DCGAN project