picture picture picture picture picture picture
Searching Business Method Patents, or Dumpster Diving in G06f 17/60

A recent decision by the USPTO Patent Appeal Board on broadening the scope of business methods patents just made life a bit harder for patent searchers. Previously business method patents were embedded in software, but now removal of the "technological arts rejection" means that a business method does not have to be related to a computer or other electronic device.

Patent searchers in the field of business methods have a range of patent classifications to search in and around the field of computer programs that were useful; now since business methods may go beyond software we are back to square one until the patent classification systems catch up.

Searching patent classifications was the traditional method of patent searching in the days when there were patent "shoe files" in the search rooms for manual searching. In the search room there may have been a hundred or so patents in a given patent classification "shoe" - carefully organized pigeonholes for paper copies of patents that grew slowly over time.

By comparison, the general International Patent Classification for business methods G06F17/60, defined as "Digital Computing or data processing. Administrative, commercial, managerial, supervisory or forecasting purposes" contains over 125,000 inventions internationally (pending or issued), according to Derwent.

The new 8th IPC edition in 2006 has a slightly improved breakdown of useful subclasses, but unfortunately the entire back file for these 125,000 records will probably not be re-classified on most databases. (The new code to replace G06f17/60 will be G06Q; IPC version 8 is here). The explosion in software and computer related patents has always made patent searching in these areas extremely difficult because the US and international classification search tools could never keep up.

Many of these 125,000 records are for Patent Cooperation Treaty applications that have been abandoned since the dot.com frenzy of a few years ago and are spinning in limbo as prior art. A log of these PCT applications emanated from the USPTO, whose staff is legendary in their ignorance of the International Patent Classification. Their errors in classification spread like a virus as other patent offices such as the Canadian Intellectual Property Office accept this PCT data without checking or verification in national phase documents.

Business method patents are not just a made in the USA situation - many are filed in Canada through the PCT system. Officially the Canadian Intellectual Property Office does not allow business methods patents, but the most cursory browsing will reveal dozens of examined and issued Canadian patents on financial transaction systems and other business methods, usually held by major banks. An example is a recent filing by the Bank of Montreal on a "Modular Financial Services Instrument" (CA 2428962, classified in G06f17/60). The BOM application asserts claims to a system for providing a multifaceted financial instrument involving credit cards, lines of credit, mortgages, deposit accounts and also methods of applying these elements.

Some PCT apps. have been indexed in databases such as Espacenet by the European Patent Office using the excellent European Patent Classification (ECLA, an IPC offshoot) which provides for an broad array of subclasses for business method patents. However, how many can be searched through ECLA is an open question, since the EPO refuses to examine business method patents which it views as "non-technical."

Patents and patent applications published by the USPTO of course are indexed by the US patent classification, the major class for business methods being 705. However, a business method patent could be cross indexed in 707 or 709 or any of a dozen other classifications dealing with computers, communications, graphics, displays, telephony, etc. The theory at the USPTO was the classification for a given patent should reflect the subject matter of all the claims. It was not unusual for a US patent to be cross-classified in 10 or so subclasses, which was very good practice by the USPTO.

However, given that some of these beastly biz method documents have hundreds of claims and exceed 500 pages often this level of diligence disappears. As a matter of fact, anyone reviewing currently published US patent applications will notice that often the document has a single US classification (and an incorrect IPC classification, or sometimes both are wildly off the mark.)

The answer to this problem is full text keyword searching, using massive search scripts of dozens of synonyms (from the generic to the minutely descriptive) in proximity with more arrays of search terms to produce every permutation or combination of terms that could produce a meaningful result. Since huge data sets usually result, one must set up filters, restricting search results to ball park classifications or to claims or to controlled index terms. Finding anything relevant is a huge challenge, given the variations in technical terminology, bad translations, cunning use of patent jargon and the abuse of language in allowing a patent applicant to be his or her own lexicographer . (I think patent agents using the term "virtual" in claims every time they draft a patent deserve a special spot in hell.)

As a general strategy a searcher needs to map her search terms into arrays such as the following:

((generic terms) or (specific narrow technical terms) or (patent jargon/acronyms))

(Near N words)

((generic terms) or (specific narrow technical terms) or (patent jargon/acronyms))

(Near N words) ..etc etc.

For instance searching for a web based pay parking reservation system may produce a preliminary search strategy like this:

(web or internet or network or LAN or wireless or cellular or mobile or Bluetooth) (5n) (kiosk or POS or point (2n) sale or ATM or self?(2n) terminal or vend?(2n) machine) (5n) (pay(3n)parking or vehicle(2n)storage)

Subsets of search strategies may be created to produce better results by bringing generic terms in closer proximity, or expanding the proximity for narrow, specific technical terms. If results are displayed in keyword in context format quick decisions can be made for record relevance.

Large data sets of results may have to be filtered by a set of patent classifications, beginning with G06f 17/60 to exclude large numbers of garbage hits. Searches need to be re-run several times by tweaking proximity operators and other search elements to verify results. Often the shortlist is still over 300 records which must be crawled and checked one by one.

Obviously one needs a powerful and fast text searching interface that can handle dozens of simultaneous search terms with a good search command language, such as the Lexis Nexis suite of full text patent databases or similar Dialog or Questel databases. Preferably searching is done at a fixed price and not on an hourly rate, and includes many double espressos to keep you awake.

The bad news is that the search resources are not improving. National patent offices are experimenting with machine classification and other cost cutting measures. Database vendors offer us goofy bells and whistles rather than improving the quality of the data, or the search engines, or the retrieval tools, or the help desk response.

I have some extra printed copies of my poster presentation "Tips for
Searching Business Methods Patents in Full-Text Dialog Databases" at NE PIUG Workshop October 16, 2001 , at ExxonMobil, Clinton , NJ and can mail them to those who is interested.

In this presentation I discussed how to combine broad and narrow terms
(concepts) using different proximity operators (including obscure DIALOG
operator (nX), or exclusive (nW)) to develop targeted strategies for business method patents. Your feedback is welcome.

In most circumstances, the best tool for patent research is the European Patent Office Classification (ECLA) accessible from the EPO Espacenet site. EPO staff have taken the International Patent Classification and broken it down into finer classification that are extremely useful for patent searching.

The catch-all IPC class G06F 17/60 "name here" including business methods patents includes over 100,000 documents

While Espacenet covers patent data from most jurisdictions, the application of ECLA coding appears to be on only documents treated by the EPO. . edition of the IPC covers 70 000, while ECLA covers 129 200

Rem: has next ed of the ipc struck??? that Canadian Patent Office does not reclassify PCT applications entering the Canadian national phase? What's wrong with that, you may ask? IPC classification of PCT applications is assigned by the receiving office. If it happens to be the USPTO, there is an estimated 50% chance that the application will be misclassified, sometimes badly (examples provided upon request). This makes classification searches pretty much useless, not only for published PCT applications, but also for Canadian PCT transfer applications, in view of the CIPO policy not to interfere with any screwed up classification. I love ECLA and use it in my searches any time I can do it, but it will remain meaningful only when coming from a single source, as is presently the case, i.e., when assigned by EPO examiners. By all means, apply it to Canadian or any other applications, but protect its integrity. I hope this is what you had in mind when talking about adopting ECLA for Canadian and US applications. I would be horrified seeing a US examiner assigning ECLA to any application, knowing his/her "fluency" in IPC as proven by PCT applications classified by US Receiving Office (not to mention the fact that ECLA codes frequently differ from IPC codes assigned even by the EPO). As a practical matter, however, if assigned by a single source ECLA cannot be a primary classification for either Canadian or US applications, or, for that matter, even PCT applications. IPC will likely have to remain the primary classification, with all the above horrors.



Home | Sitemap | Contact Us
Last Updated: January 6, 2008