An experimental NLG system which uses GF WordNet to generate articles about entities in Wikidata.
The advantage of the approach is that GF WordNet already provides a lexicon for >20 languages with an abstract API of >100k words. Of those ~30k words are linked to their Wikidata items. On top of this core lexicon, I integrated ~600k given and family names plust ~4 million geographical names from Wikidata. This means that the total lexicon contains close to 5 million lexical items most of which are linked with their corresponding Wikidata items. These are the starting point to generating articles about people, countries, capitals and cities.
For the NLG we don't need to write an application grammar. You can use the RGL api and the lexicon directly from Python. Look at nlg.country, nlg.capital and nlg.city for examples.
There is also a mock up web interface here:
https://cloud.grammaticalframework.org/wikidata/index.wsgi
Try searching for countries and cities! The response is currently slow since it takes time to fetch the data from Wikidata. This is something to be improved in the future.