I saw PatentMojo a while back, but I didn’t mention it because it came so close to being right that the fact that it was wrong pissed me off too much. But now I see that O’Reilly Radar has picked it up and it is getting lots of interest, so I feel obliged to explain myself.
I like that PatentMojo has taken a messy database and turned it inside-out, exposing the data as feeds. That’s the right place to start. And adding Feedburner subscriber data is smart too, as it is a way of exploiting the community to find out what’s worth watching.
(Sidenote: But the site also inadvertently shows how community/collaborative recommendations can be screwed up by community characteristics. Case in point, PatentMojo shows 84 people subscribed to Apple’s patent feed, while zero people are subscribed to Qualcomm’s feed. Well, Apple earns virtually no money from its patent portfolio, while Qualcomm earnings close to half its income from patents. Go figure.)
So what don’t I like? This is still first-gen stuff. There is no entity extraction. It’s all keyword-driven. Sure, I can track Apple, say, but can I keep an eye on patents that reference Apple? How about UI patents? How about patents that touch on areas where Apple has announced interest? Can I, in other words, track things at an ontological and conceptual level, as opposed to purely at a keyword level? No, so PatentMojo ends up dumb and backward-looking, not smart and forward-looking.
The trouble with these first-gen RSS-tracking tools is that they don’t do anything more than commodity keyword-tracking. They will be roadkill as soon as decent second-gen tools come along that understand how to extract interesting entities from unstructured data.