In universities and colleges around the world, Google has encouraged the creation of "Google Developer Student Clubs". To round out the year, Sam Witteveen and I decided to offer our three-part "Back-to-Basic" series to clubs in the South-East Asia region, to introduce them to TensorFlow and Deep Learning.
This session was the third in the three-part series, and focused on Deep Learning for Natural Language Processing using Transformers (which is unusual for an introductory course, but is more 'current' than choosing the standard 'WordEmbeddings+RNNs' approach).
Our presentation was in two parts:
My initial talk, that :
- Gave a brief outline of the use-cases for Transformers for NLP tasks
- Introduced two new Layers : Tokenisation and Transformer
- Outlined the BERT task, and how we could then apply Transfer Learning
Sam did a Code-Along using Google Colab, where he lead the audience through :
- Loading and preparing an App review dataset
- Building a TensorFlow (Hugging Face) BERT Classifier
- Fine-tuning the model on our dataset, and looking at the outputs
The slides for my talk are here :
If there are any questions about the presentation please ask below, or contact me using the details given on the slides themselves.