Classified Information: Patent Searching by Classification Part 1

Classification searching is the original method of patent searching. Physically housed in “shoe boxes,” copies of US patents are organized by class and sub-class at the US Patent and Trademark Office in Alexandria, VA as well as Patent and Trademark Depository Libraries located around the country. Prior to digitization of US patent documents, the only way to search this patent collection was by browsing through the various subclasses by hand (or by eye, through microform).

Today’s post will highlight why classification searching is still necessary and how to go about finding relevant classifications. Next week I will briefly touch on some advanced classification searching methods. For a detailed and thorough course on classification searching as well as many other patent searching methods and techniques, check out Patent Resources Group’s Art and Science of Patent Searching course. This comprehensive three day course takes place from April 11-13 in Bonita Springs, FL as well as from August 16-18 in both Alexandria, VA and Southfield, MI. Designed for patent attorneys and agents, inventors, paralegals and research managers, the course and its materials will be understood by anyone from the inexperienced to advanced practitioners.

Classifications, by definition, are groups of patent documents of similar subject matter. Several major patent classification systems exist, including US, IPC, ECLA, Japanese F-Index and F-Term, and DEKLA.

With the advent and prevalence of keyword searching, most people default to punching a few words in and hoping for the best. This mentality in our every day searching lives can transfer over to patent searching, when we want the best and most relevant art in the fastest way possible. Who would manually flip through tens, hundreds, or thousands of patents when a carefully constructed search string input into Thomson Innovation or PatBase delivers results right away? Keyword searching is not without faults, however. First of all, keyword searching is limited by the digital text data available within the system. The most widely available full-text collection of US patents only dates back to 1976, excluding nearly four million patents published prior to that date. Secondly, searchers are limited by the choice of words themselves. Searchers must try to think of every relevant keyword possible to include in the query, which may be difficult especially when a term is general. What is called a “protrusion” in one case may be called a tongue, nub, nodule, bump, or even the dreaded “member” in another document. Since the patent applicant is his or her own lexicographer, there is almost no limit to what the desired feature may be called. For these reasons, and several others, classification searching is still important to today’s searcher.

In order to search by classification, one must first identify relevant classifications. There are a number of ways to do this including: using classification directories, looking on the face of relevant patents, and utilizing statistical analysis.

Classification directories, as I will be referring to them here, are listings of classifications and subclassifications by definition of subject matter. The USPTO Classification Home Page contains a searchable class schedule, US to IPC or Locarno concordance, and browsable indexes of the classification structure. WIPO’s IPC directory is searchable by IPC code or text. Intellogist offers Best Practice articles divided by subject matter that often include key classification areas, such as Chemical Engineering and Electrical Engineering. Starting out a search by going directly to the classification directory is advantageous because no keyword searching needs to be done up front. One noticeable downside is that classification areas can be hard to understand if one isn’t familiar with not only the subject matter being referenced, but the classifications themselves. Deciding between A61F 2/14 (Prostheses implantable into the body.. Eye parts, e.g. lenses, corneal implants) and its child class A61F 2/16 (…Intraocular lenses) can be difficult without prior experience and no patents to compare and contrast.

For this reason, one good technique for identifying relevant classifications to further investigate is by looking on the face of a patent of interest for the classifications that have been assigned by the patent examiner. If I know that US patent 4,883,485 A is closely related to the prior art I am trying to locate, I can look on the face of the patent and see that it has been classified under IPC A61F 2/16 and begin to look in that classification more narrowly.

A great way to expand this technique is to first find several patents that are related to the prior art desired (through methods like keyword searching or citation searching) and then statistically analyze the batch to find which classifications are most prevalent. PatBase, Thomson Innovation, Qweb, QPAT, and MicroPatent PatentWeb are a few examples of patent search systems that offer analysis features including classification analysis. For example, PatBase allows users to perform classification analysis on US, IPC, ECLA, and F-Term classifications and Thomson Innovation has many different ways to display the classification analyzation results. It is important to be selective in choosing the patents prior to classification analysis. If the subject matter is too broad, the statistical noise generated may obscure the best classifications.

In the second installment I’ll touch on some advanced techniques for classification searching, but if this blog entry has intrigued you at all, you may be interested in Patent Resources Group’s Art and Science of Patent Searching course. This comprehensive three day course takes place from April 11-13 in Bonita Springs, FL as well as from August 16-18 in both Alexandria, VA and Southfield, MI.

Feel free to ask any questions in the comments or share your own classification tips!

Read Part 2.

Patent Resources Group

This post was contributed by Intellogist team member Chris Jagalla.


22 Responses

  1. Social comments and analytics for this post…

    This post was mentioned on Twitter by Pandersonpllc: Great post @Intellogist, too many people don’t know how to do this.

  2. Hi
    I know you’re primarily looking at classification searching and not patent searching as such, but whenever I do pharmaceutical searching I back up a broad keyword with a classification symbol (or vice versa) so you catch those applications that have either been slightly misclassified, and those that have odd (including misspelled) keywords.

  3. Insomniac,

    I agree with your strategy. In fact, in a post coming up this week I’ll be detailing how I combine keyword searching with classification searching to wring out those useful patents.

  4. Often several keywords are used in multiple industries with different meanings, or the same meaning with different applications. Either way, you don’t want results for automobile engines if you’re looking for cellular pH sensors. In this case using the EC classification search [ G01N33/68F] with the word sensor will allow you to bring up a manageable group of protein sensors [35]. This eliminates anything that may have the word protein and sensor in the same abstract/title [742]. The caveat being, you are at the mercy of the person who classified the application and no two people classifies things the same way, when a ENT sees an elephant he sees a trunk, when a proctologist sees the elephan he’s just a big…

  5. I definitely find that classification searching is one of my top means to find a reference, along with inventor/assignee searches and citation searches of other important references. Text searching usually works the best when I’m looking for a part of a reference that is very specific.

    As an aside, I remember taling to a professor of engineering technology as month or so ago, and he mentioned that classification searching has been his main method of searching for the past 30 years.

  6. Mike
    Your strategy is an appropriate way to restrict the size of a search although I would use a truncated classification symbol such as G01N33* because you should expect that the classifier is sufficiently skilled to get the right class even if they don’t get the same final subclass as you. You may have a few more to look at but you are not so much at the mercy of the classification.

  7. […] Patent Searching by Classification Part 2 Posted on March 24, 2010 by cjagalla Last week I introduced patent classification searching and gave reasons why it is still relevant in today’s keyword-centric search environment. […]

  8. Hi Christopher,

    Thanks for such a nice and informative post.

    I have a small question about patent classification searching.
    I understand that patents are classified by examiners based on the subject technology. Generally, such patent classification is done at (very) early stage of examination. Separately, various patent classification systems regularly update (Revise) the classes and subclasses and definitions thereof.
    I want to know –
    Do the patent offices have any procedure in place to reclassify a patent that was earlier misclassified or if the class definition has changed?
    If yes, then do various search engines (Micropat, Google patents, and the like) take care of such reclassification of patents?


  9. Great question Abhishek!

    Unfortunately, there is not “one” answer to your question. Certain classification systems, such as the IPC, do revise their classification systems and retroactively apply these classifications to older patents. Please see ( for more information on how they do this. It is worth noting that the Core/Advanced section of the IPC article on Intellogist was correct at one point, but due to the introduction of the IPC version 10 earlier this year, it is now out of date–we will be updating the article soon.

    Likewise, the USPTO does initiate revisions to the USTPO class code on a per-class basis – my understanding is that this is based on requests from examiners and other PTO staff. After the classification manual is revised, all patents in the affected classes are retroactively re-classified. Because of this, patents that were misclassified would not necessarily be identified and fixed.

    As far as patent search systems, most advanced for-pay search systems such as PatBase and Micropat incorporate these re-classification efforts into their data files. Google Patents, however, does not. This is because their database is built on scans of the patents themselves–what you see on the front of the patent is what you get, including whatever classifications were issued at the time. Some of the problems with Google Patents, including this classification issue, are touched upon here (

    Abhishek, does this answer your question? Let me know if there’s anything I can do to be more specific–and check out part 2 of this article on our home page!


  10. Automated reclassification can’t work where a number of new classes have been expanded from a single old class (unless of course there is some semantic engine thingo that can try and match the subject matter to a new class) so they should be manually reclassified, and I don’t know if they are. Years ago at IPONZ we had to reclassify from IPC 4 to IPC 5 manually because we weren’t fully computerised then, and it’s not a pretty job. Espacenet and PatBase provide an IPC 8/9/10 class as well as an IPC 1-7 class so if you happen to search the new class or old class you still pick up all relevant publications.
    Also at IPONZ we used to classify when the application was made, and then reclassify at acceptance.

  11. Thanks for sharing your experiences insomniac!

  12. I just wanted to add one more note about the USPTO’s re-classification process in light of insomniac’s comment. My understanding is that the re-classification is actually done by hand. I thought I heard somewhere that this is done via a contracting agency, but can’t find a citation to prove it.

  13. Thanks Chris, this did answer my question(s) to the entirety :)

  14. […] by classification works and why it’s worthwhile. To check out those posts, see parts one and two of the “Classified Information” […]

  15. If I had a buck for every time I came to Superb read.

  16. Thanks for the compliment Corey!

  17. […] Classified Information: Patent Searching by Classification Part 1 […]

  18. […] Classified Information: Patent Searching by Classification Part 1 – This post really took off in part because there were several followups on the topic that covered an essential part of patent searching: different classification systems. […]

  19. […] Better classification data is important for better classification searching, which is an important part of any searcher’s […]

  20. […] there are many many other techniques for using classification in a search. We’ve written about this subject several times on the Intellogist® Blog (and I’m quite sure this […]

  21. […] Classified Information: Patent Searching by Classification Part 1 – A primer on how to search patents by classification, along with why it’s so important! […]

  22. […] laid out above for cases when the area of focus is clear prior to entering the search. Preemptively combining keywords and classifications is a particularly powerful search strategy to mitigate and navigate through the growing mass of […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: