This is interesting....I am trying to come up with a way to create, for a given intent, X sentences through NLG that communicate the same intent. For example, to say "What are you grateful for?" but have the NLG ensure that if asking the same question every day, no two days in successful would be the same by relying on NLG.

Nice write up, on an interesting subject, Fabio! These days I was preparing a course about computing for the humanities, and I was looking at Python's Natural Language Toolkit http://www.nltk.org/.

BTW, the correct sentence in plural form would be "There are some finished and delivered stories that *were* not tested." Natural languages are tricky...

Btw, I just checked out the main page of NLTK and it seems like it's an NLP (a language parser/tokenizer) instead of an NLG (generator). Is there a generator module as well?

I never used NLTK for generating text. I don't know that is your goal with text generation. I've only done it for fun, and for that purpose a Markov chain generator works suprisingly well for such a naive algorithm. I tried to find how to generate text with NLTK and apparently the feature is temporarily disabled...

BTW, there are multple books about NLTK, and the official docs were also published as a book by O'Reilly in 2009.

Got it, in my case it's not a "random" generated text through learning such as with a Markov Chain, but a deterministic, programmatic template to precisely control how I want my phrases.

You're right, I thought that, but no grammar check triggered warnings so I posted, but it sounds wrong, so I changed the complementary phrase to be more correct and a bit more polite as well. Thanks.

Your post is also a beautiful illustration of how a DSL can make coding in Ruby so elegant. I guess that was your main point. Great post.