As a CSIRO 2018-2019 Summer Vacation Scholar, I built a deep learning model for analysing the emotional content of social media content in three commonly used languages within Australia: English, Chinese and Arabic.

Initially I used scikit-learn to build an SVM classifier, but eventually moved towards building on Google's then-recently released BERT deep learning architecture using the PyTorch Transformers library (then known as pytorch-pretrained-bert).

Analysing emotions in tweets using deep learning

A multilingual emotion detection model trained on social media texts, based on Google's BERT.

Analysing emotions in tweets using deep learning

Project description