In Q1 2008, Jim Sinur and I will release research on the different approaches to business rule representation. This is going to be a fun research effort – lots of theory, lots of hands-on testing. This is also a misunderstood area. For instance, some users think that a natural-language (NL) approach to rule representation is the only way to go. NL is great and has its targeted applications – which we will discuss in the research – however, NL is no panacea. This is where those same people get concerned and say, “Hey! Don’t we use natural language all the time? Isn’t that what you’re writing in right now? How can there be anything wrong with natural language?”
My usual research response has been to remind you that your tax code and all your laws are written in natural language (“oh… I see your point. I never could understand that deduction for the profits on excess mineral rights that can be claimed by part-time fishermen who install solar panels on their second homes in July.”) No, I won’t go there. Instead, let me tell you about some tires I bought.
The tires are great! My wife and I love them. And we thought we loved the financing scheme the company gave us. Here is what it said on the contract:
“6 months no interest/minimum payment required”
Wow! Six months and we don’t have to pay interest or make a minimum payment. I like that rule. Well, we threw away the first bill and when we got the second bill, there was a charge for failing to make the minimum payment. Come again? I called the bank behind the promotion and they said the rule should be read as (a) 6 months no interest but (b) a minimum payment is required. This is the natural language pitfall – semantic interpretation.
Folks, where I come from – a slash tends to be read as an “or” and distributes the verb uniformly. This is equivalent to an airport warning, “no knives/handguns allowed.” No one in his/her (I had to do it) right mind would interpret that to mean: (a) no knives but (b) handguns are allowed. Try explaining the logic behind that interpretation as you wait for your orange jumpsuit and a new roommate at the nearest prison.
In a rare stroke of customer service luck, the agent immediately agreed with my interpretation. Perhaps he is a closet linguist. He waived the fees and in return, I am just going to send in the whole payment and pretend this event never happened. I love those tires. I no longer love the financing scheme.
In defense of NL, if you allow poor or ambiguous grammar you are asking for trouble. The NL rule tools do a great job at clarifying the ontology and enforcing proper grammar. And NL has its place. But my point is simple. If you think natural language is a universal panacea, you are in for a surprise. We learned too much from the days of AI to forget that.
Gartner clients: watch for our research in Q1. NL is just one of the rule representation approaches we will be reviewing in detail.