I’ve spent seven years developing gnowit.com. It arose out of my PhD thesis, and employs the codebase and architecture I put together over the years.
Gnowit.com contains some very novel innovations around sentiment classification, auto-categorization of information, core-topic extraction and ‘noise-cancelling’ (let’s leave it at that for now).
As we’ve come out of stealth mode two weeks ago, I’ve found that it has become important to share details of these with potential partners, who may actually be in the same field, and would find them valuable.
The way to approach this is to either get an NDA (which is ineffective when the partner already has related IP in the same area) or to get patents. Thus, I am forced to take the second route.
There are some interesting wrinkles when it comes to software patents (and patents in general). They have to use a certain language, and for software in particular, you have to take pains to emphasize that it is not an abstract operation that is being carried out, but one where the ‘signals’ are being transformed. You actually need to write the patent as if you’re taking a lump of rubber and producing something from it!
There is an entire art-form related to the claims, where you try and ‘capture’ as wide a claim as possible, and yet, keep the claims attached to the concrete processing (so they do not get dismissed for being too vague or abstract).
It’s like writing a research paper, but with one hand tied behind your back, and without the need to prove anything (except that you can actually build something useful from this !).