The following tutorials show how to create a NLP classifier that achieves 84% accuracy on the classification of Yelp store reviews with 738 training examples.
This repository consists of the following:
1-Construct-TensorFlow-Data-Pipeline.ipynb - This is a Jupyter notebook created in AWS Sagemaker that shows how to create a data pipeline in Tensorflow.
2-Sentiment-Classification-with-BERT.ipynb - This is a Jupyter notebook that takes information created in 1-TFRecord and uses it as the basis of a NLP classifier. This notebook utilizes Huggingface's Transformers library and a language model called BERT.
3-How-to-use-Saved-Model.ipynb - This is some example code that shows how to utilize a TensorFlow SavedModel.
bert_config.json - This contains the parameters that configure the Huggingface library.