Previously we’ve identified the twin Achilles heals of NLG adoption as 1) the fairly common reluctance of traditional writers to embrace NLG, and 2) a hubristic belief among technologists that, FINE, we can build an NLG system without you writers. A third problem might be that NLG vendors position the technology as something that’s good for the business in the long run rather than immediately useful and profitable right now. “Make better decisions,” “improve data literacy,” and the like, sound about as immediately beneficial as taking your daily vitamin.
Nila June: Instant Property Descriptions gets beyond all three of those potential problems by taking the NLG output directly to the consumer. The premise is simple: a user (presumably a real estate agent with a home to describe) answers a series of questions about the property, and Nila June instantly returns a property description that is suitable for publication on home listings sites (Zillow, Trulia, Realtor, Homes, and a host of others).
The four primary ingredients of any NLG recipe are 1) structured data, 2) subject matter expertise, 3) writing talent, and 4) the technology to pull it all together. Nila June brings numbers 2 and 3 naturally. Using AX Semantics, Nila June also brings number 4. But what about number 1? Structured data is as important to NLG as eggs, flower, and sugar to a bakery. Well, Nila June gets these critical ingredients from the customer through an interview sequence, and then presents the customer with a finished cake.
Nila June is a good example of what we mean when we say that the road to commercially deployable natural language generation is straighter than might first appear.