Sentence Analyzer - enabling business and enterprise applications to handle sentences and text

Yes, it has been around since Mr. Chomsky, Mr. Pinker and many many others. Volumes of research, and there must be 100-200 APIs out there for various things. But of late, in how many business APIs have you seen them being actively used ? What could be the reason why business products avoid them ? One of course is the adding-machine mentality, the database bias. But another strong reason I think it boils down to - ease of use.

XML and Java, not Prolog : Many of the existing APIs seem to favor Prolog, I think a lot of language research was done when Prolog was popular. I love Prolog too, but technology seems to have an arrow of time, for example, a cute Prolog-Java interpreter I was tracking for years suddenly vanished without trace. The key thing is, modern business applications are built by people who know more about Java and C#. This toolkit is built on Java. More important, this toolkit emits XML and strings, and anyone can use it with very little ramp-up.

Look ma no math, only some customizable grammar rules : There is a heavy element of statistics and maths used in most linguistics packages. I do not like statistics being applied to a bag of words too early in the process, so I have used none here. If the results compare or exceed or enhance current APIs, then surely adding the math later will be even better, so I think this is the right approach.

Grammar experts in business apps ? : I know literally nothing about Chomsky's theories, in their details. But I am sure some people do. This demo has lots of options for injecting high level grammar rules, tiered or not. This toolkit makes it possible use the domain knowledge of grammar experts in business applications, to bring many of their theories to practical implementations of great use.

Highly flexible : There are two layers in this demo. One, uses the Universal Grammar. The other uses commonly used English Grammar. This dual layering opens many possibilities. It may be possible to replace the English Grammar layer by , say Scandinavian grammar (and the dictionary too, naturally). What the results may be I do not know, but if Universal Grammar holds, it should work fine. I also expect it to work fine for languages like Chinese, which I know has very structured sentence construction rules.

Moral of story : Comparing this to the Stanford NLP POS tagger is not a good idea, the results could be inconclusive or surprising.

More reading for those interested...