Post by Janet Freilich
Ted Sichelman has written a response to my earlier post on whether there are too many patents to search. He argues that the problem of patent searching is smaller than I suggested. He includes an excellent critique of Christina Mulligan and Timothy Lee’s article on patent search costs and provides some much needed numbers on the magnitude of search time and costs. I agree with him on many points, but I stand by my conclusion that clearance searches are very difficult, if not impossible.
Sichelman notes that it only takes a few minutes to review a patent, so conducting a search and then reviewing a large number of results is not too onerous. He is correct that it will often only take a few minutes to determine whether or not a given patent is relevant, however, my sense is that the real problem is not only in reviewing the patents you find in the first search (which may get you 95% of the way to clearance), it’s in finding the last 5%. Those are the patents that use non-standard vocabulary or otherwise cover the product in unexpected ways. You have to slog through a lot of unrelated results and useless searches in order to be completely confident that you haven’t overlooked any of the results. Patentees can famously be their “own lexicographers” and “redefin[e] the meaning of particular claim terms away from their ordinary meaning” as long as the term is defined elsewhere in the patent. Merck & Co., Inc. v. Teva Pharms. USA, Inc., 395 F.3d 1364, 1370 (Fed. Cir. 2005).
For example, if you wanted to know if you could freely manufacture a table, you could search for patents claiming the word “table.” However, a patent could accurately describe a table as a “horizontal surface having at least one vertical leg.” If you didn’t know beforehand what language was being used, how would you find that patent?
The problem with patent searches is that if you miss even one patent that covers your product; you don’t have legal clearance to manufacture. Different industries differ in how thoroughly they require clearance searches, and my sense is that thoroughness is generally more important in the life sciences.
Sichelman also notes that there is technology being developed to help with patent searches. This is a fascinating and welcome development, and it is important that the problem be tackled from both technological and legal angles. I am eager to see the interaction between legal technology and legal rules discussed more in the academy. I am also curious to see how the law, especially malpractice law, adapts to patent search technology.
I am not aware of malpractice litigation relating to patent search technology, however, a case relating to a manual patent search is instructive. In Carabotta v. Mitchell, 2002 WL 42948 (Ohio Ct. App. Jan. 10, 2002), the client sought a patent attorney’s opinion on whether a specific type of umbrella could be manufactured in the United States without patent infringement. The attorney’s clearance search returned approximately 1000 patents, which the attorney “flipped through,” scanning the title page and drawings. Although the attorney’s search returned a patent covering the client’s umbrella, the attorney did not notice that it was relevant and advised his client that he had not found a patent covering the umbrella. The client eventually realized he did not have clearance to manufacture and sued the attorney for malpractice. The outcome suggests that opinions on the question of malpractice were split. The jury found for the attorney, the trial court granted a motion for a new trial, commenting that the attorney’s “efforts fell well below the standard of care,” and the appeals court reversed and remanded with instructions to reinstate the jury’s verdict.
E-discovery might provide a guide. There, the overproliferation of documents led to technological development but also novel legal questions and types of malpractice suits. As an increasing number of documents were stored electronically, and written electronic communication increased, the number of documents retrieved during discovery became unmanageable. Furthermore, the standards for gathering and storing electronic information were far from clear. In response, technology companies developed methods for sorting through electronically stored information at issue in litigation. Also in response, the Federal Rules of Civil Procedure were amended to ease the difficulties posed by electronically stored information.
I’ve been reading this debate with great interest. Well, I’m not sure it is a “debate” as much as a disagreement over degree. But it seems that both Janet and Ted agree that the costs of patent clearance are more than they should be if one wishes to have confidence in their freedom to operate. It is also difficult to agree on generalizations when, as both authors point out, there is significant variance across industries in willingness to conduct clearance searches in the first place.
Google’s attempts at automation are valiant, but I’d be leery of any sort of artificial intelligence being applied to claim coverage, at least at the present time.
The true solution is one that represents a legal innovation as much as a technical one. Sullivan and Lee suggest that only chemical patents can be”indexed” and as a result carry far less clearance costs. However, this is only true under the conventional searching paradigm, which I refer to as “positive searching.” With positive searching, you request the subject matter that you wish to see.
But if we adopt a system that utilizes “negative” or eliminatory searching, then any technology can be “indexed” to reduce clearance costs at least as effectively (and likely more) as achieved by STN with their chemical structure databases. The basic idea here is that patent claims are associated with an index of technical elements that are required for infringement. Claims can be broken down into one or more distinct elements, each of which is a required aspect that must be present in a product in order to infringe the claim.
To perform the clearance, the searcher navigates the index of technical elements (which is in hierarchy form) and selects the concepts that are not present in the product. Patents that are associated with the selected elements are removed from consideration without the searcher needing to set eyes on the patent itself. Patents that claim the same elements are removed all at once, increasing efficiency tremendously.
Janet also mentions the problem with finding patents that use different terminology or describe aspects in unexpected ways. This system takes care of that too because the indexing process effectively results in terminology normalization. Moreover, extremely broad or out-of-the-woodwork patents are not missed because of the eliminatory nature – patents are only removed from consideration based on conscious and deliberate decisions as to required elements. If a patent is not eliminated, it is returned as a result.
This is what we are doing at ClearstoneIP.
Gabe,
Thanks for the insightful thoughts. I still think AI can be very useful in automated searching, but I agree that a conceptual/indexing approach is an important ingredient in reducing search costs, and I look forward to reviewing your materials on your website. If you have any technical papers, feel free to post a link, of course.
Thanks, Ted. I certainly agree that AI is helpful generally in patent searching, but in the specific context of FTO it tends to result in too much noise and misses key patents (either those that are very broad or that use unexpected terminology/description).
The question at issue in FTO is fundamentally different from that in patentability/prior art searching, so we need to be operating under a different searching paradigm. An FTO method should be able to avoid patents we don’t want to see (those whose claims require something not present in the target product), not return a huge pile of patents based on disclosure. In other words, the focus should be on the claims, not the disclosure.
Here are two of our recent blog posts demonstrating how this different approach plays out:
http://www.clearstoneip.com/blog/?p=63 (Five Reasons to Memorialize Claim Analysis)
http://www.clearstoneip.com/blog/?p=73 (Bridging the Divide between Patent Analysts and Engineers)
To quantify the efficiency gains, we are talking about getting to the 5-10% most critical patents in an indexed set (containing, e.g., several thousand patents) in less than an hour.