How to add non text data to a text classifier?

Suppose you more data from your IMDB reviews beyond the pure text, like the gender, age and localization of the reviewers.

How would you integrate that info into your classifier? How would it change the training process?

1 Like

You can create a vector which contains extra features apart from text and you can concatenate to the text vector. But the common approach would be to add extra features in the later layers of the network. There are times when it can give improvements.

1 Like

Where can I find some references? Iā€™d like some guidelines about doint it.