Data-driven Natural Language Generation systems use machine learning techniques to learn how to produce natural language utterances from a given input.

Template-based text generation makes use of one or more templates, either hand-crafted or learnt from training data. Inputted data is then inserted into one or more of these templates to produce an output.