Using the Cooperative Patent Classification to find early patents

The Cooperative Patent Classification (CPC) is mainly meant as a tool to assist the searching of modern patents by subject, but can also be used as a wonderful way to find many older patents.

The document CPC coverage sets out the date spans for which countries' patents are covered on the free Espacenet database., where the Classification link leads to the classification schedules. I have checked sample classes and cannot agree on the dates given for comprehensive coverage for two countries.

These are France, where coverage is supposed to be from 1844, and I make it 1902, and Switzerland, where it is supposed to be from 1939 and I make it 1920.

Below are listed the main patent authorities and their coverage:

United States, 1836 to date [i.e. from beginning of numerical series]
Germany, 1877 to date [i.e. from the first publication]
France, 1902 to date
United Kingdom, 1909 to date
PCT, 1978 to date
European Patent Convention, 1978 to date

Going back so far (over 8 million US documents alone) must have been an incredibly hard task and all credit is due to those involved. It is easy to quibble -- I'd like to see titles supplied for all of those covered by the CPC, ideally in English rather than the local language. Applicant and inventor data is also often absent. Adding such data would immensely improve the use of the Espacenet database for historians of inventions. I do realise that this is unlikely to be a high priority.

Top of the list, for me, would be to extend CPC coverage for the UK back to 1893, when the earliest patents appear on the database. Those searching are likely to use English alone, or with another language, and besides the American material it would be very useful to be able to search further back. Maybe material before 1893 could be added ?

So, how can this material be accessed ?

The CPC schedules on the Espacenet database should be used first. It can be searched by keyword to find possible classes, when the user then drills down to the detailed classes. These can then be transferred to a search mask in the database to run a search, to which additional limits such as by publication date spans and country can be added. Adding applicant or inventor names is hazardous, as often these fields will be absent.

For example, enter in the search box A42B5/00. This is the class for veils for faces, and could also have been found by entering "veils". Tick the hollow box next to the class. This enters it on the left, below which is "Copy to search form". I usually use this in preference to "Find patents" as that would find everything.

Once in the search form, limits can be added. For example, enter in the "Publication date" box 1844:1914. This limits the search to that date span. I found 69 hits. These can be sorted by priority [original filing] date, ascending order, although this may not work well for very early material, as it is often not indexed, so that the earliest appear first.

The "publication number" field can have one or more country codes such as US, GB, FR, DE to limit hits to that country. Hence US reduces the hits to just three, covering 1907-1914. One such is Veiling by Charles Gaskill of Delaware, as illustrated below.

Three sounded a low number, so I searched the title field by the word "Veil?", where the ? allows an s to be added. There were 35 from 1899, although many were for veil fasteners or brooches etc. A search by title word of the British material gave 87, though only from 1894 onwards.

Clicking on a patent title then means that the actual patent can be seen as a single PDF by clicking on "Original Patent", while clicking on "Mosaics" means that the drawings can be seen as miniature. At the top, "Next" always enables the following hit to be seen in the same format, which is useful for scanning through lots of material.

A big problem for anyone using the database is not knowing what has been indexed -- the patents themselves, or some fields, may be absent. "Absence of evidence is no evidence of absence" -- because you didn't find it may not mean that it doesn't exist. Nevertheless, searching can often reveal useful and often unique material for the history of technology.

The Adidas watch that checks your pulse: patent classification problems

There has been a lot of publicity about wearable technology that diagnoses you -- checking your heartbeat or pulse and so on.

When I first heard of this idea I thought of shirts incorporating technology, but most of the interest is in wristwatches. An example is in Adidas' new sensor for athletes, so that training programmes can be planned, and there is a BBC news story on it, Adidas Micoach smartwatch has heart-rate sensor

Now, supposing you wanted to identify inventions by Adidas in that field. You could ask a patent database for Adidas plus keywords such as watch. Better would be to take the patent classification A61B5/024, "Detecting, measuring, or recording pulse rate or heart rate."

This is an IPC -- it will be used by patent offices when they publish patent specifications. A61B5/02438 is a CPC which is a more precise class, not used on the specifications but added to those published by leading Western nations plus the World PCT system. It is for portable devices worn by the patient (or any user, really). It might seem obvious to use that class combined with the company name but it is easy to miss material, so while any detailed search would certainly use it, a second search should always be used with the broader class plus keywords.

Let's call using A61B5/024 for Adidas Search 1, and A61B5/02438 Search 2.

It is still necessary to decide which search box in Espacenet you use: CPC or IPC. The results below give first CPC and then IPC for Search 1 (Search 2 doesn't have an IPC option, as it is not a valid class).

Search 1 gives in CPC 4179 hits, with 2 by Adidas; in IPC 7127 with 5 by Adidas.

Search 2 gives in CPC 3129 hits, with 4 by Adidas.

How can this be ? The apparently broad search found fewer hits by Adidas as CPC, but more with the IPC than the apparently more precise CPC.

Besides the fact that the company may be working in related fields, and double-counting when more than one entry appears for the same invention, a problem is that using simply A61B5/024 says that only that class is wanted. Any specification classified by a more detailed class instead is missed.

Espacenet deals with this problem by allowing anyone searching the classification to enter the class in the search box which automatically selects the subordinate classes. They can then be copied to the search form. This results in a CPC search which we can call Search 3.

Search 3 gives in CPC 12,925 hits, with 7 by Adidas.

These include the intriguingly named Method and system for extracting cardiac parameters from plethysmographic signals, illustrated below.

Six of the seven are for similar coats that do the monitoring etc., while the seventh is not obviously for a watch -- it is a "portable electronic unit" which monitors congestive heart failure patients, and hence obviously not athletes.

So which is the best approach ? If asked to search for inventions for the concept by Adidas, I would use broad classes in both CPC and IPC (as the company has relatively few inventions). I would use the "In my patents list" ability (below the title in the bibliographic format) to list patent specifications by Adidas in the general field for later retrieval. I would also use the descriptions and the search reports at the end of European or World publications to see if they mentioned related material by Adidas.

I also searched the very broad A61B together with timepiece, watch and wristwatch (and their plurals) for Adidas and got zero hits (but beware their clever use of "portable electronic unit", which includes most watches).

It does seem that Adidas has not yet published the details -- hardly surprising, as it takes 18 months for a new specification to be published.