msahamed/yelp_comments_classification_nlp

Implements three progressive architectures—LSTM with standard embeddings, LSTM+1D convolution for efficiency gains, and LSTM+convolution with pre-trained GloVe embeddings—to classify 1.6M review comments as positive/negative based on star ratings. Text preprocessing includes tokenization, sequence conversion with 50-word fixed length (zero-padded or truncated), and evaluation across Keras models to identify optimal speed-accuracy tradeoffs. Built with NumPy, Pandas, scikit-learn, and NLTK for the full NLP pipeline from raw text to model training.

No commits in the last 6 months.