Hands-On Machine Learning with TensorFlow.js

Loading the IMDb dataset

The IMDb sentiment dataset is the most famous movie review dataset. It contains text that can be used to evaluate movies with a positive/negative sentiment.

We are going to use the preprocessed dataset and the code in tfjs-examples to load the IMDb dataset since it contains several util libraries that we can use to convert the IMDb dataset into a TensorFlow.js-compatible format. Please see the original code to find out how to load the IMDb dataset.

Here, we will load sentences that are a maximum of 100 words in length. The vocabulary size for this dataset is 10,000. This is the same as the sentiment analysis example in tfjs-examples:

const numWords = 10000;const maxLen = 100;const embeddingSize = 8;console.log('Loading ...