The 7 Simple Secrets to Patent Searching

Tips and techniques for searching patents

We are in the middle of a revolution for technological supremacy.  A new age of patent rights has arrived marked by ever-expanding court battles over patent claims with billions of dollars at stake.
 
The role of the patent searcher has radically changed as well.  Today, a cursory search, which may have been adequate in the past, may put your company or client at risk.  The good news is that the systems and software for navigating patent data have greatly improved, as have the techniques for using them.
 
This article examines how some of today’s most successful patent searchers approach their job.  In it, I reveal 7 simple secrets that patent searchers can apply to perform a world-class patent search. 
 

Patent Search Tip 1:  Use Patent Classification Search

Classification searching is really easy.  It is often more accurate, far quicker and more comprehensive than keyword searching.  It just requires a little practice. 
 
Today there are four major classification systems for patents.  The US has its own (USPC), Europe has another (ECLA), WIPO and most of its member states use another (IPC), and Japan has one (F-term).  Each is unique, and each has its own approach to patent taxonomy.
 
The good news is that starting in 2013, the US and Europe will be transitioning to a unified classification system called the CPC (Cooperative Patent Classification system).  The CPC will be based on ECLA, which is itself based on the IPC.  
 
As a result, there will be, in effect, two global patent classification systems (The CPC is a superset of the IPC), and my guess (I have no insider information) is that Japan will adopt the CPC once it is released and settles down a bit, but they may not abandon F-terms entirely because F-terms uniquely classify patents by their features which is a handy way to search for Japanese patents.  F stands for “feature.”
 
Since there are more than 150,000 unique subclasses in the US system alone, some patent searchers will cross reference an index to find the relevant classes, but there is a much quicker technique:
 
I am trying to find patents for tire locking devices like the “Denver Boot.”  Since I don’t know much about these devices, I searched for “Denver Boot” because it was a term I am familiar with, and I figured that somebody would use the string in his or her patent.  I came up with just 9 US grants, but the best one was classified in 70/226, which is a three-dot subclass described as follows:
 
Class:  LOCKS
Mainline Subclass:  RSnubbolt
One-Dot:  For control and machine elements
Two-Dot:  Wheel
Three-Dot:  Rotation Blocking
 
Clicking into that sub-class I find 112 patents, and if I extend my search to children subclasses of 70/226, I get 267 US grants.  I found 267 US grants, all of which are in some way like the Denver Boot, in only 15 seconds.  There is no way you could do that using keywords.
 

Patent Search Tip 2:  Query by Example

Query by example is a search technique that lets you search for documents in the patent corpus by using an existing known patent as an example or “seed document.”  You can then ask the patent search engine, “Show me more documents like this one.”
 
The most common methods to query by example use the language or keywords in the patent text.  The algorithms create a complex multi-term query designed to capture a very large set of documents and sort the most similar ones to the top of the list.
 
Automated techniques for finding similar patent documents can be enhanced with synonym dictionaries and, in some cases, Latent Semantic Indexing (LSI), which requires a pre-computed lexicon.  
 
Alternative methods use a patent’s classification fingerprint (a fingerprint of the original and cross reference classes, or inventive and non-inventive classifications) in a carefully crafted machine-generated query that finds documents that are classified “like” the target patent.  I personally find that class fingerprinting gives better results than even very sophisticated linguistic techniques because it is not limited by the vagaries of human language.  However, I still use both methods, because they complement each other.
 

Patent Search Tip 3:  Cluster for Themes and Concepts

Patent searchers often dismiss clustering software because they feel the assertions by software developers overstep the true capabilities of clustering.  While this is certainly true, it doesn’t mean clustering has no value to the patent searcher.
 
Clustering organizes large sets of documents into logical themes or concepts.  
 
Consider the 267 documents in class 70/226 (and children) that were like the “Denver Boot.”  After clustering the set, I get a primary theme, “Wheel Lock,” and several sub-themes including “Lock Assembly,” “Tire,” “Locking Rod,” “Bicycle,” “Locking Pin,” “Locking Bar” and more.  
 
cluster-results
 
Remember, I don’t know much about these devices, but now I have three new terms that I can use in my search that I may not have thought of:  “Locking Rod/Pin/Bar.”
 
You can think of clustering as a Keyword Extraction Tool, not for one patent, but for an entire set.  Use clustering and your keyword-based search strategy will be vastly improved.
 

Patent Search Tip 4:  Use the Boolean OR operator 

You already know basic Boolean logic.  There are three Boolean operators you use when patent searching:  AND, OR and NOT.  The tendency for most searchers is to overuse the AND operator because ORs bring back too many results, but this is undeniably the wrong approach.  
 
To illustrate, say I have 20 terms, including synonyms, which describe the invention I seek.  If I use the AND operator, I’ll get just a handful of hits, or even zero hits, because the chances of one document containing all 20 terms is low, particularly when some of the terms are synonyms.  
 
In the “Denver Boot” example, the patent drafter is likely to use either “Pin,” “Bar,” or “Rod,” but not all three terms.  So naturally you’ll start deleting terms, but now your recall is diminished, and you’ll miss patents that contain the words you deleted.
 
A better approach is to join your terms using the OR operator.  You will likely get over a million documents that contain at least one of the terms in your list.  Don’t let that intimidate you.  
 
Of course you are not going to read a million documents; you don’t have to.  The secret is in the sorting algorithm.  If you sort by relevance the best matches across all 20 terms will appear at the top of the list.  You’ll see this right away.  As you page through the results, at some point the patents will no longer be relevant, but who cares?  Ignore them.  The point is the best matches will be at the top of your list!
 

Patent Search Tip 5:  Use Patent Data Value indicators

There are well-known and academically studied value indicators in patent data.  The most studied example is the number of forward citations a patent receives.  
 
When a patent application is filed, patentees must submit a list of prior art references called backward (reverse) citations.  The patent examiner adds to this list during prosecution.  
 
Each patent cited by the patent application receives a forward reference in turn.  That is, a newer patent document cites it as an example of the prior art.  Patents that accumulate forward citations at a high rate are more likely to be important or fundamental patents.
 
As a result, once you have narrowed your set using some of the other techniques, it is a good idea to sort it by forward references, and make sure you have read and understand all the highly cited patents.
 
Other value indicators include:  number of inventors, length of claim one, size of simple patent family and even number of drawings/figures.  
 

Patent Search Tip 6:  Iterate Your Searches

The chance of finding all relevant patents in a single query is about as likely as Apple and Samsung settling their lawsuit because they want to stay friends.  
 
In the Denver Boot example, I started with a keyword search, then did a classification search, then clustered the result set to look for new themes, then narrowed by more keywords.  I then read a few patents and found some that were spot-on, and used two different query by example techniques to find the other patents that were most like that one.  I found other patents that were relevant and repeated the process using those patents as new seeds.
 
Good patent search is a manic process.  You’ll zoom out with a broad search, get a feel for the results, then zoom into a few classes, assignees, or dates, then zoom in further to specific patents, then zoom out again based on information you learn about the patents that that you find.
 
Zooming in and out allows you to take detours and uncover all the relevant art related to the goals of your search!
 

Patent Search Tip 7:  Use Charts During Investigation Process

Charts are used most often to present your final results to your management or client, because they give a high level overview of your findings.  However, they are just as useful during the discovery process.  
 
For example, if you have narrowed in, more or less, to the relevant set, it is a good idea to chart by assignee or chart by date so you can see both who are the main players (assignees), and when the technology was at its early stages, reached its peak, and started waning.  
 
Charts clearly display trends and aid your search.
 

Conclusion

There’s no doubt that the future belongs to patent holders.  As technology gets more interwoven and must borrow from those who have come before, a strong patent portfolio may be the only doorway into the marketplace.
 
During the last few years, patent searchers who have helped their companies and clients gain competitive advantage have become indispensable members of the team. Advanced software is part of this trend, and any patent searcher who is expert in the use of the high-end platforms is that much more valuable.

Post navigation

LATEST ARTICLES