Sentiment Analysis NLP for QMIND

A machine learning program that maps typed text to emotion, built with an innovation design team at QMIND, Canada's largest undergraduate AI organization.

Approach

• Dataset: Reddit comments labeled with one of 27 emotions from Google's GoEmotions.
• Preprocessing: tokenization, stemming, lemmatization, and stop-word removal.
• Model: a fine-tuned ALBERT classifier that predicts the most prevalent emotion in a sentence.

My role centred on researching candidate ML models and sourcing the data that shaped the final solution.