README

In order to train the model, one must execute the notebook 'Auto encoder.ipynb'

This notebook expects to have glove embedding 200dimensions in a subfolder
'glove/'.

The notebook will generate a file checkpoint.h5 in the '/tmp/' folder. This file
will be used for further scripts.

To execute predict.py the usage is:
python3 predict.py 'Tweet content in simple quotes' anger
where anger is the main sentiment of the tweet.

To execute the prediction on a large batch of tweets:
python3 predict_all.py data.csv

Non-standard requirements are tensorflow, keras, numpy, jupyter and pandas.

The notebook offers the possibility to use tensorboard to monitor the model
training process, the logdir is under /tmp/encoder