Yesterday I spent part of the afternoon at a US Patent & Trademark Office roundtable discussion on using crowdsourcing to improve the patent examination process. Thanks to Chris Wong for looping me in and helping to organize the event. If you’re interested, you can watch the whole video here.
I was there not as an expert in patents, but as someone who represents lots of small startup internet companies facing patent issues, and as someone who spends a lot of time on the problem of how to solve challenges through collaborative processes (basically everything USV invests in).
Here are my slides:
And I’ll just highlight two important points:
First: why do we care about this? Because (generally speaking) small internet companies typically see more harm than benefit from the patent system:
And second, there are many ways to contemplate “crowdsourcing” with regard to patent examinations.
In the most straightforward sense, the PTO could construct a way for outsiders to submit prior art on pending patent applications — this is the model pioneered by Peer to Patent, and built upon by Stack Exchange’s Ask Patents community.
The challenge with this approach is that while structured crowdsourcing around complex problems is proven to work, it’s really hard to get right. A big risk facing the PTO is investing a lot in a web interface for this, in a “big bang” sort of way (a la healthcare.gov), not getting it right, and then seeing the whole thing as a failure.
To that end, I posed the ideas that getting “crowdsourcing” right is really a cultural issue, not a technical issue. In other words, making it work is not just about building the right website and hoping people will come. Getting it right will mean changing the way you connect with and engage with “the crowd”. As Micah Siegel from Ask Patents put it, “you can’t do crowdsourcing without a crowd”.
We also talked about the importance of APIs and open data in all of this, so that people can build applications (simple ones, like notifications or tweets, or complex ones involving workflow) around the exam process.
Tying those three ideas together (changing culture, going where “the crowd” already is, and taking an API-first approach), it seems like there is a super clear path to getting started:
- Set up a simple, public “uspto-developers” google group and invite interested developers to join the discussion there.
- Stand up a basic API for patent search that sites like Ask Patents and others could use (they specifically asked for this, and already have an active community).
That would be a really simple way to start, would be guaranteed to bear fruit in the near term, and would also help guide subsequent steps
Or, to put it in more buzzwordy terms:
It felt like a productive discussion — I appreciate how hard it is to approach an old problem in a new way, and get the sense that the PTO is taking a real stab at it.
After you proposed a simple API, the question from the USPTO folks was “What is an API?” These are smart people, but their expertise is in patent law, not building software.
Crowdsourcing patent examinations is a good idea, but I think there is lots of low-hanging software fruit the USPTO could address before crowdsourcing. quick example – reading their official manual is a pain, so I ripped the data and published a readable copy via github pages.
official – http://www.uspto.gov/web/offices/pac/mpep/s1134.html
readable – http://adlervermillion.com/mpep/s1134.html
Hardly the biggest software problem, just the easiest to fix from the outside ;-)
Yep
BTW that is great
You should join the developers group when it opens up!
Yep
BTW that is great
You should join the developers group when it opens up!
—
https://ngis.lndo.site on the fly
I’ve written about this in the past because it’s a problem for the PTO, patent lawyers and basically anyone analyzing patent issues: http://www.patdek.com/blog/2011/06/23/crowdsourcing-your-invalidity-contentions-and-prior-art-searches?rq=crowd
A takeaway from that post – information analyzed should be easily reusable and not be discarded, but that’s not what happens today.
Maybe a better example is looking at what AskPatents (StackExchange) is offering. I once tried to put together a quick response to show what was wrong with a patent under review. This turned into a project spanning an hour. Even though I knew the prior art well, and it was an easy task to line up against the model claim, the process was still fairly cumbersome. Take a look: http://patents.stackexchange.com/questions/5224/mobile-search-organizing-search-results-based-on-user-location-and-availabilit/5242#5242
Mainly because I had to type information about the prior art into the system or cut and paste it. And once I put that analysis into AskPatents, someone would have to copy what I put on the site to reuse it. The analysis is left static, flat. Not terribly usable by a patent examiner, but certainly better than just dumping the prior art on an examiner.
What can be done? Snippets. Capture and save image excerpts (and the text within that image). Dump that snippet information from the document in a database. Because the snippet is really what matters. So gathering patent content and creating snippet tools for PDF documents is where we’ve focused our efforts with Patdek.
I’m always happy to talk to anyone about what the process is to truly analyze patent claims, so we can all make this an easier, more efficient process.
Yep, totally agree
They need http://documentcloud.com for prior art
—
https://ngis.lndo.site on the fly
We’re very familiar with their stuff and what CrocoDoc was doing before the Box purchase. The DocumentCloud open source is a good example to draw from. An enthusiastic group of developers backed by supportive funding. They have some neat and simple approaches to snippets.
To make crowdsourced prior art more extensible during examination, a deeper set of tools would need to be added to the DocumentCloud stack. There’s definitely a lot people can understand from the code, though.
I’m one for two on AskPatents – Google’s “friends night out” patent was rejected (http://patents.stackexchange.com/questions/5182/tracking-and-managing-group-expenditures-google-patent-application-prior-a)
But HP’s overboard Semantic Web patent was not (http://patents.stackexchange.com/questions/4668/semantic-web-policy-enforcement-hewlett-packard-issued-patent-prior-art)
IMHO, the HP one was more important as it may affect the future of the agent-based Linked Web.
Anyway, if patent applications are purposely written to maximize obfuscation (the exact opposite of an elevator pitch), then unless that changes, we’re just nibbling at the edges of the problem.
Patent applications are just not engineering documents, which they were when the Patent Office started out. It’s actually doing more harm than good and fundamental change is required.
Maybe USV can lead by example by fostering a new kind of patent application for its portfolio companies? Or is that tantamount to unilateral disarmament?
BTW that is great here’s definitely a lot people can understand from the code, though.
You should join the group when it opens up to sohbet chat developers! Continue up
BTW that is great here’s definitely a lot people can understand from the code, though.
You should join the group when it opens up to Chat Odaları developers! Continue up