Session Ready
Exercise

Classifying news articles

In this exercise you will create a multi-class classification model.

The dataset is already loaded in the environment as news_novel. Also, all the pre-processing of the training data is already done and tokenizer is also available in the environment.

A RNN model was pre-trained with the following architecture: use the Embedding layer, one LSTM layer and the output Dense layer expecting three classes: sci.space, alt.atheism, and soc.religion.christian. The weights of this trained model are available on the classify_news_weights.h5 file.

You will pre-process the novel data and evaluate on a new dataset news_novel.

Instructions
100 XP
  • Transform the data present on news_novel.data using the loaded tokenizer.
  • Pad the obtained sequences of numerical indexes.
  • Transform the labels present on news_novel.target into a one-hot representation.
  • Evaluate the model using the method .evaluate() and print the loss and accuracy obtained.
Loading