Report:Google Patent Search/Search Interface/The Search Forms/Limitations on Number Searching

Limitations on Number Searching in Google Patents

Users who are familiar with other patent-industry directed search engines will notice that there is no "number search" bar on this quick search form, for users to input document numbers. Likewise, the Advanced search form only offers a search field for granted patent numbers, and no way to search by other document identifying numbers. Because the Quick Search bar searches all scanned textual data (including bibliographic data), it effectively functions as a number search feature. Users should note the following quirks of the system, however:

• Six-digit granted patent numbers (e.g.5,656,655) and eight digit application numbers (e.g. 10/155,083) can be entered with or without punctuation.
• Users cannot include the country code "US" in front of any of these numbers, as would be customary in many multi-national patent search systems. Including "US" will cause the search to return no hits. Document kind codes will also cause this result.

Google Patent Search uses the application number (also called the "application serial number"), rather than the publication number, to identify US published applications, as of mid-2008. This is apparent in the screenshot below.

The engine uses eight-digit application numbers to identify US published applications as of mid-2008.

Another quirk of the system is that both application numbers, and publication numbers, are omitted from the electronic bibliographic data of granted patents (though they can be seen by reviewing the document front page). As seen in the image below, only the issued patent number is shown in the electronic bibliographic data.

Google Patents does not index publication or application numbers for granted patent documents.

Through empirical testing, it seems that while a search on the former publication number (with correct punctuation) will retrieve only the granted patent, a search on the issued patent's application number will retrieve only the corresponding published application, and fail to report the existence of the granted patent.

A search on the original application number retrieves only the related published application, while a search using a publication number finds only the related granted patent.

This result comes about because publication numbers seem to be no longer indexed as bibliographic data, but are often listed on the front page of granted patents and are found through a pure text search. On the other hand, application numbers seem to be specially bibliographic indexed on US published applications, but not on granted patents, and a text search doesn't seem to be able to find them.

Editor's Note:

The decision to use application numbers to represent the US published application collection is going against the accepted protocol used in most search engines. One reason for this is that it destroys the distinction made between granted patents and published applications.

As background, application numbers (also called "application serial numbers") used to consist of six digits, and when the number 999,999 was reached, the serial number would begin again at 000,001. Later, the US patent office began adding a two digit "series code" to the beginning of these numbers, to distinguish a series 1 serial number from its identical series 2 counterpart. Thus, a typical application number (or serial number) will have a format of, for example, 10/167,125. Unfortunately, the reason they do not make good document identifier numbers is that in the earlier years of the patent office, these two digit series codes were not used/not available, and a six digit serial number without its code can refer to multiple patent documents.

The developers' decision to use application serial numbers to identify applications is also somewhat strange because it theoretically destroys the distinction between applications and grants: even granted patents contain serial numbers. In other words, patents issued pre- March 2001, although they have no corresponding published application (because the US did not issue published applications before March 2001), will still have corresponding application serial numbers. However, this problem has apparently been solved by Google Patent Search because the system only indexes published applications by application serial number, and does not index the application serial numbers of granted patents. As a result, sorting a hit list by "application number" does effectively show only actual US published applications, and eliminates granted patents.