Module:Sandbox/Jeblad/NLG/doc explained

The inputs to a NLG-system will depend on a 4-tuple

\langle k, c, u, d \rangle

Where $k$ is the knowledge source (Wikidata), $c$ is the communicative goal, $u$ is a user model, and $d$ is a discourse history.

Steps in the processing are

• Content determination
• Document structuring
• Aggregation
• Lexical choice
• Referring expression generation
• Realization

Not all steps will be done as usual as there are several simplifications that can and will be done.