Hi Martin, I’m afraid that I don’t remember. It seems plausible, though. You might try generating a noun from its sigma projection as a simple test. Cheers, John From: "Forst, Martin" <Martin.Forst@nuance.com> Date: Saturday, August 31, 2019 at 3:35 PM To: "xle@mailman.uni-konstanz.de" <xle@mailman.uni-konstanz.de>, "Maxwell, John <maxwell@parc.com>" <maxwell@parc.com>, Ron Kaplan <ron.kaplan@post.harvard.edu> Subject: generation from s-structure (rather than f-structure) Hi Ron and John, hi all, I’ve written a small grammar with an s:: projection from the f-structure. Parsing works as expected, and so does regeneration as long as I don’t add or remove features. However, my ultimate goal is to generate from s-structures, so I’ve manually reduced some files resulting from parsing to just the respective s-structures and tried generating from those. Even though I’ve made all f-structure features addable, generation fails. (And so does regeneration when I make that same list of features removable and addable.) Am I trying something that the generator isn’t made for (yet)? Thanks, Martin On October 1, Nuance Automotive will become Cerence Inc. – A.I. for a world in motion. [cid:image002.png@01D5494F.B08DBA60] Dr. Martin Forst Principal Research Application Scientist “Cer” pronounced like air, “ence” said like the word sense. A new word for a new company. Born from Nuance Automotive, Cerence fuses intelligence, calm and control. Exactly how your journey to the future should feel. cerence.com<http://www.cerence.com>
