Creating your own custom machine learning model to classify text

Using the default interface to classify text based on sentiment or general categories is very powerful but doesn't allow us to classify different texts based on our own rules. Being able to create our own model is where the full power of machine learning lies.

Fortunately, Google offers the power to create and train our own models based on our own set of training data. This allows us to generate a collection of texts and classify them using our own labels. With this data, we will prepare our own model that can be matched against new texts.

We will see in this recipe an example of classifying emails sent to a shop that has two sections, "appliances" and "furniture." We will create a third category of "others" that should capture emails that don't fit neatly into either category.

The process is highly dependent on the quality of the data that is provided to the model...