coming out of a discussion with Joshua Montgomery of Mycroft, I’ve put together a tool for building natural language training data. http://opennludata.org
Currently it reads and writes JOVO/Mycroft and RASA format files.
Offers the option to publish skills into the github repository under an MIT license.
Only a sample of skills at this stage, hoping some folks out there will share their training data. Big diverse training data sets are key to good ML models.
Hope it’s useful