Text Classification App

Challenge

Text classification has a wide variety of applications, such as hate-speech detection, sentiment analysis, or general purpose categorization of medical, legal, or engineering documents.

In this project we used a deep learning algorithm based on BERT to classify a corpus of 1.8 millions text pieces into 7 categories. The computation was performed on a 8 GPU AWS instance to speed up the process. After training, the model was operationalized on an AWS Multi Model Server.