U.S. patents have gotten a lot of press lately, with the Supreme Court’s expected spring-time 2010 decision in the case of Bilski v. Kappos likely to determine whether future business process methods (think Amazon’s one-click checkout) can be patented. That subject has gripped the attention of the software industry, where some worry that a positive decision could potentially halt development of new technologies, create costly infringement lawsuits, and hurt competition. Others, of course, may worry that a negative verdict will impact the profits they can make through licensing their patented business methods. That can apply to legitimate inventors, but also to the patent trolls who are in the business of buying up patents mainly to profit from licensing them -- without themselves having contributed to or making use of the innovation in any way -- and aggressively litigating against infringements.

But software vendors aren’t the only ones who could have reason to keep a closer eye on U.S. patents and patent applications and their own patent portfolios. There are a number of reasons companies in every industry may need to be able to research these quickly and efficiently – and an upcoming API to enable access to a semantic similarity search matching tool from TextWise can help them when it comes to identifying intellectual property opportunities and risks. Now in a private beta service, it lets users search about five million U.S. patent records to find a ranked list of semantically similar patents that can help them, for example, determine whether a troll claim they face is likely invalid, screen new patent filings, discover new portfolios or patents they may want to acquire, or ferret out potential licensing and sales efforts for their own patents and portfolios. With its semantic similarity search matching technology, organizations should be able to quickly find non-obvious opportunities in the patent and application portfolios, TextWise says. “This will cut down tremendously on the amount of work that it takes to actually work with patents in the future,” says TextWise CEO Connie Kenneally, and the vendor’s goal is to make it easy for all employees to efficiently mine IP data.

The current database covers U.S. issued utility patents and utility applications from 1978 onward, “It’s a huge amount of data,” says Connie Kenneally, and the company has plans to expand it further over the coming months, such as possibly including foreign patents. “Think about it even in a research setting like pharmaceuticals, where we think there will be a big use, where they are looking to patent new ideas and concepts. All these patents or applications for patents pile up on someone’s desk. Who’s going to try to screen them?” she says. TextWise’s system can come up with all patents and applications that are similar to them so that now there’s a smaller number of things an employee has to investigate without having to search the entire database himself. “We bring back a limited list, if you will, that might qualify for further investigation,” she says.

Patent searching is a rather well-developed if labor-intensive science that today demands experienced patent search experts who aren’t in huge supply at many companies, Kenneally notes, so it’s been interesting to bring semantic similarity searching technology into this mature space. The feedback, happily, has been beyond what TextWise expected, she says. “Think of the semantics behind this—it’s to a point where you can take either a patent number, an entire patent or any piece of text of that patent and now match it to any U.S. patent or application that has come out in the market since the ’70s,” she says. “That’s a really hard thing to do because patents are quite detailed. When you start talking about taking a 30-page document, or patent, and getting key portions of that patent in a similarity search and have it match with very highly relevant ranked results, that’s pretty interesting.”

Now patent researchers don’t have to take on the broad-based cuts of where to look for things that might be similar to what their company is considering patenting, for example. Or when a patent troll comes in to make a claim against technology, or perhaps even business methods around an IT-enabled process they’ve deployed, “you have a tool that is really good at determining how close something is to something else. That’s what similarity search does; it gives you a relevance ranking that tells you Piece A is similar to Piece B.”

One of the companies that has looked at the technology was also able to use it to distinguish among different product approaches as they appear in patents, she says, saving a lot of time in narrowing down the lists of potentially similar approaches. Says Kenneally, “We’re just beginning to learn where it all fits.”