Understanding topic classification

We saw a small example of topic classification in Chapter 5, Text Classification and Regression Using AutoKeras, with the example of the spam classifier. In that case, we predicted a category (spam/no spam) from the content of an email. In this section, we will use a similar text classifier to categorize each article in its corresponding topic. By doing this, we will obtain a model that determines which topics (categories) correspond to each news item.

For example, let's say our model has input the following title:

"The match could not be played due to the eruption of a tornado"

This will output the weather and sports topics, as shown in the following diagram:

Figure 8.1 – Workflow of a news topic classifier

The previous diagram shows a simplified version of a topic classifier pipeline. The raw text is processed by the classifier and the output will be one or more categories.

Later in...