Attorney, Agent or Firm: the <agent> element

Some IP firms have multiple offices. Those firms’ internal docketing systems undoubtedly include detailed particulars of each patent case prosecuted by the firm, with an indication of which one of the firm’s offices is responsible for each case. Multi-office firms can accordingly use data mining techniques to develop metrics representative of various aspects of the operations of their different offices. However, that is feasible only for those with access to the data, i.e. the specific firms which have accumulated the data internally.

Can data mining techniques be applied to bibliographic patent data to develop metrics representative of the operations of different offices of a multi-office IP firm? Not directly.

Consider United States Patent No. 8332851 which issued on 11 December 2012 to SAP AG for an invention of Ostermeier et al. entitled Configuration and Execution of Mass Data Run Objects. The red-underlined portion of this partial image of the ‘851 patent’s cover sheet tells us that the corresponding US patent application was prosecuted by the well known IP firm Fish & Richardson P.C. As of this writing, Fish & Richardson P.C. has offices in Atlanta, Austin, Boston, Dallas, Delaware, Houston, Munich, New York, Silicon Valley, Southern California, the Twin Cities and Washington, DC. Which one of Fish & Richardson’s 11 US offices handled the prosecution of the application from which the ‘851 patent issued?  The cover sheet does not tell us.

Now consider this small extract from the <us-bibliographic-data-grant> element of the USPTO’s XML publication for the ‘851 patent. The <us-bibliographic-data-grant> element encapsulates the <parties> element, which in turn encapsulates the <agents> element (as well as the <applicants> element). The <agent>, <addressbook>, <orgname>, <address> and <country> elements encapsulated and sub-encapsulated by the <agents> are sparsely populated. We see “Fish & Richardson P.C.” encapsulated by the <orgname></orgname> tag pair, but none of the other elements encapsulate data apart from the "attorney“ attribute in the <agent> element and the “unknown” text encapsulated by the <address><country></country><address> hierarchical tags. Is that “unknown" an aberration or a data glitch? No it is not.

This query result set is based on a database I constructed from the USPTO’s bibliographic data for US patents which issued in 2012. As can be seen, the rightmost “country" column (which corresponds to the aforementioned <address><country></country><address> hierarchical tags) contains the text “unknown” for every patent listed here. The same is true for the entire dataset. It is thus apparent that the bibliographic data does not tell us which one of the 11 US offices of Fish & Richardson P.C. handled the prosecution of the application from which the ‘851 patent issued. So which office was it?


The answer is the Dallas, Texas office. This is revealed by looking the case up via the USPTO’s public Patent Application Information Retrieval (i.e. public PAIR) system. Specifically, this portion of the filing transmittal for the application from which the ‘851 patent issued appears on the letterhead of the Fish & Richardson P.C. Dallas, Texas office (and also cites the firm’s Minneapolis office address; presumably to facilitate centralized docketing for all of the firm’s offices).

In summary, the USPTO’s bibliographic data identifies the “attorney, agent or firm” by name only. No address information (not even a county identifier) is provided, so it is not possible to discriminate between different offices of the same firm solely by reference to the bibliographic data.

Correlation of USPTO Art Units with IPC subclasses

What correlation, if any, is there between the art units to which the USPTO allocates patent applications for examination and the International Patent Classification subclasses allocated to the claimed inventions encompassed by the resultant patents? One would expect a fairly strong correlation, given the USPTO’s allocation of US patent classifications to art units and further in view of similarities between the IPC and US patent classification schemes.  For background, see the USPTO’s tabulation of US Patent Classes Arranged by Art Unit and its US patent classification to IPC concordance.

This scatter plot shows, along the horizontal axis, numbers of US patents which issued in 2012 for different USPTO art units; and along the horizontal axis, numbers of those patents in terms of their primary IPC subclasses; both of those measures being referenced to the assignee dimension (i.e. the parties to whom the patents issued).

The positive slope trend line reveals a strong positive correlation, as expected. The underlying trend model has an r² value of 0.843717, so r (the correlation value, in accordance with the Pearson correlation) is 0.92. Correlation values close to either +1 or -1 represent strong positive or negative correlation respectively between the measures being compared. Correlation values closer to zero represent weaker—or an absence of—correlation.

Office specific data elements

WIPO’s ST.36 Standard recommendation for the processing of patent information using XML includes a provision for so-called office-specific-data elements. These elements can be used to encapsulate details unique to a particular country in patent bibliographic data published for that country. ST.36 also provides alternative mechanisms for accomplishing the same thing, e.g. mixing office-specific elements with international common elements.

Let’s consider an example. United States patent no. 8260535 issued on 4 September 2012 to Bombardier Recreational Products Inc. of Valcourt, Québec, Canada for an invention of Mario Dagenais entitled "Load sensor for a vehicle electronic stability system". The ‘535 patent issued from United States patent application no. 11/864,265 which was filed on 28 September 2007.

As of this writing, an apparently corresponding Canadian patent application is pending, namely CA 2699332 which was published by the Canadian Intellectual Property Office as of 2 April 2009. The ‘332 application is a Canadian national counterpart of international application PCT/US2008/070129 which was filed on 16 July 2008 and claimed priority based upon the US ‘265 application. PCT/US2008/070129 was published on 2 April 2009 as WO2009/042276.

Consider this small extract from the <ca-bibliographic-data> element of the Canadian Intellectual Property Office’s XML publication for the ‘332 application. The <ca-bibliographic-data> element encapsulates the <publication-reference> element, which in turn encapsulates the <document-id> element. The <country>, <doc-number>, <kind> and <date> elements encapsulated by the <document-id> element identify this as Canadian patent application no. 2699332 published 2 April 2009, as aforesaid.

Now consider this corresponding small extract from the <us-bibliographic-data-grant> element of the USPTO’s XML publication for the ‘535 patent. Some structural similarities are evident: the <us-bibliographic-data-grant> element encapsulates the <publication-reference> element, which in turn encapsulates the <document-id> element. The <country>, <doc-number>, <kind> and <date> elements encapsulated by the <document-id> element identify this as United States patent no. 8260535 issued 4 September 2012, as aforesaid.

The foregoing are examples of international common elements.

Now consider this small extract from the <ca-office-specific-bib-data> element of the CIPO’s XML publication for the ‘332 application. Among other things, this encapsulates the <ca-license-for-sale> element.  This is a uniquely Canadian feature whereby a patentee may request that a patent be flagged as available for license or sale when details of that patent are published in the CIPO’s weekly Canadian Patent Office Record. In this case the <ca-license-for-sale> element encapsulates the word "false", indicating that no such flag has been raised in respect of the ‘332 application—not surprising since the ‘332 application has not issued as a Canadian patent as of this writing.

Finally, consider this small extract from the concluding portion of the <us-bibliographic-data-grant> element of the USPTO’s XML publication for the ‘535 patent. Here we have some uniquely American features identifying the primary examiner (Khoi Tran), the USPTO art unit (3664: robotics and vehicle controls) and the assistant examiner (Bhavesh V. Amin) for the ‘535 patent.

We can thus see that the CIPO has opted to implement ST.36 by assigning office-specific-data elements to encapsulate details unique to Canada, whereas the USPTO implements ST.36 by mixing office-specific elements with international common elements.

Plant patents

The US is one of the few countries that grants plant patents. Canada grants "plant breeders’ rights" which are administered not by the Canadian Intellectual Property Office but by the Plant Breeders’ Rights Office, which is part of the Canadian Food Inspection Agency.

The USPTO’s group art unit 1661 handles plant patent applications. US plant patents are allocated one of two kind codes: P2 or P3; depending on whether the application from which the patent issued underwent pre-grant publication (P3) or not (P2).

The USPTO’s bibliographic patent data contains a wealth of information for plant patents, including the botanical denomination (i.e. Latin name) of the genus and species of the patented plant; and the plant variety designation (i.e. cultivar name) of the patented plant. For example, US plant patent PP23241 issued on 4 December 2012 for a plant having the botanical denomination Echinacea purpurea and variety designation "Quills and Thrills". According to the patent’s abstract, the plant is "characterized by large inflorescences with quilled ray florets of purple pink, a compact, multicrown habit, a long bloom time, and excellent vigor".


It’s relatively straightforward to feed bibliographic data of this sort directly from a visualization tool into a search engine, to obtain further information. For example, the botanical denomination or the variety designation can be fed into a search engine directly off the visualization to get an image of the plant—as seen here for "Quills and Thrills".


As further examples, either the botanical denomination or the variety designation can be fed into a search engine directly off the visualization to lookup the corresponding plant patent(s) in the Google patents database, or to see if the botantical denomination appears in The Plant List—as seen here for Echinacea purpurea—etc.



Click the "Plant Patents" tab above to explore these and other aspects of the USPTO’s 2012 plant patents.

The USPTO’s patent bibliographic data—concatenated XML

In a previous post I mentioned that the Canadian Intellectual Property Office’s 2012 XML format patent bibliographic data is provided in a single 188 MB archive from which 58,572 separate XML files can be extracted. 21,592 of those files correspond to Canadian utility patents which issued in 2012. The other files correspond to Canadian laid-open applications (kind code A1), reissue patents (kind code E), re-examined patents (kind code F) and republished versions of previously published files (e.g. to correct errors). The files range in size from about 2-39 KB, with an average size of about 3.3 KB.

In contrast, as shown in this portion of Google’s "USPTO Bulk Downloads: Patent Grant Bibliographic Data" web page, the USPTO’s 2012 patent bibliographic data is provided in 52 separate .zip archives—one per week (recall that US patents are issued in batches, on Tuesday of each week throughout the year). Unless you relish the prospect of manually downloading 52 separate archives one at a time, you’ll want to consider using a bulk file download utility.

After downloading one or more of the USPTO’s weekly bibliographic data archives you can extract the contents of each archive. Unlike the CIPO’s archive, which extracts into a multiplicity of XML files—one per Canadian patent bibliographic document—the USPTO’s archives extract into just three files. This example shows details of the three files extracted from the USPTO’s bibliographic data archive for week 4 of 2012, i.e. bibliographic data for US patents which issued on Tuesday, 24 January 2012.

The .txt file contains a checklist of the patent numbers included in the archive, with one patent number per line. For week 4 of 2012 the checklist includes design patent numbers D0652606 through D0653015, plant patent numbers PP022464 through PP022468, reissue patent numbers RE043120 through RE043146 and utility patent numbers 08099794 through 08104093. The .html file includes a header (shown here—click to enlarge) summarizing the contents of the archive.

Notice that there is only one rather large .xml file. This is a concatenated XML file. As explained in the USPTO’s Bulk Data Product FAQs:

It is important to understand that the concatenated XML documents in the ZIP files, which have file extension "XML," are not the same as standard XML files and therefore will not be immediately readable by an ordinary XML parser. Instead, the files must be broken into individual XML documents, by splitting them apart at the XML declarations and/or DOCTYPE declarations.

Thus, unlike the CIPO’s archive from which one may directly extract separate XML files corresponding to individual Canadian bibliographic patent documents, some further processing of the USPTO’s concatenated XML files is required. Since XML files consist only of text and since each separate XML document within the USPTO’s concatenated XML file is prefaced by a unique XML declaration header (e.g. <?xml version="1.0" encoding="UTF-8"?>) it is relatively straightforward to split the concatenated XML file into separate XML files. For week 4 of 2012 this should yield 4,725 separate XML files as shown in above header.

Working with XML format patent bibliographic data

The USPTO issues United States patents in batches, on Tuesday of each week throughout the year. The Canadian Intellectual Property Office does the same: Canadian patents are issued in batches, on Tuesday of each week throughout the year.

According to the USPTO’s statistics 253,155 US utility patents were granted in 2012. I’m ignoring reissue, design and plant patents for comparison purposes. Canada does not grant design or plant patents. Instead of design patents, Canada grants industrial design registrations. Instead of plant patents, Canada grants plant breeders’ rights (these are administered not by the CIPO but by the Plant Breeders’ Rights Office, which is part of the Canadian Food Inspection Agency).

A search of the CIPO’s online patent database reveals that 21,592 Canadian utility patents issued in 2012. So, in 2012, the volume of Canadian utility patent grants was about 8.5% of the volume of US utility patent grants. An even greater disparity appears in relation to reissue patents: the USPTO granted 822 reissue patents in 2012, but only 20 Canadian reissue patents were granted in the decade spanning 2001-2011.

The USPTO and the CIPO publish bibliographic data for their respective granted patents in XML format, in accordance with WIPO’s ST.36 standard. The CIPO’s Canadian patent bibliographic data XML files are typically provided in .zip type archive files. For example, the CIPO’s 2012 XML format patent bibliographic data is provided in a 188 MB archive from which 58,572 separate XML files can be extracted. However, those XML files pertain not only to granted utility patents (kind code C) but also to laid-open applications (kind code A1), reissue patents (kind code E) and re-examined patents (kind code F).

Moreover, the CIPO may republish a patent bibliographic data XML file—if an error is detected in a previously published version thereof.  For example, the CIPO’s 2012 patent bibliographic data archive includes an XML file for Canadian patent no. 2121906 which issued on 29 April 1993. As shown here, that XML file contains a pair of <ca-date-updated></ca-date-updated> XML tags encapsulating the 31 December 2012 date on which the CIPO republished its XML bibliographic data file for the ‘906 patent (New Years Eve 2012 fell on a Tuesday).  In processing the CIPO’s patent bibliographic data, one must take any such republication into account and perform appropriate update operations on existing data.

This brief discussion touches on only some issues that one must be cognizant of in processing XML format patent bibliographic data. Next week I’ll discuss another issue specific to the USPTO’s XML format patent bibliographic data.

viewing XML format patent bibliographic data

OK—you have obtained some bibliographic patent data in XML format from one of the available sources (see last week’s post). Now what?

You can inspect individual XML files with an XML file viewer. XML files consist of text only, so a text editor such as the Microsoft Windows Notepad utility will do. Here is a small portion of the USPTO’s XML file for US patent no. 8329177 as viewed in Notepad. You can see the descriptive tag pairs (e.g. <document-id></document-id>) encapsulating the information content, but the tags’ hierarchical structure isn’t readily apparent via Notepad.

The tags’ hierarchical structure is more apparent if we inspect the XML file with a spreadsheet program such as Microsoft Excel. Here is a small portion of the same XML file (i.e. US 8329177) as viewed in Excel.  The file has been converted into the familiar spreadsheet row/column format, with the tags appearing as column headings and the encapsulated information shown in rows beneath the respective headings. The conversion (or "flattening") process repeats information in some cells, as seen here.  The column headers have been narrowed to show more columns.   When viewed in Excel, the XML file for US 8329177 has 53 rows and 131 columns, so the worksheet has a total of 6,943 cells. However, the worksheet is rather sparse: only 2,408 of those cells contain information.

A web browser can also be used to inspect an XML file. Here is the same small portion of the XML file for US 8329177 shown above in Notepad, as viewed in Microsoft Internet Explorer. In this case, the tags’ hierarchical structure is made apparent by color highlighting and by hierarchical indentation levels. The encapsulated information is bolded.  This makes it somewhat easier to browse through the contents of a single XML file—if that is what you want to do.

None of this is very helpful, unless you are only interested in a particular XML file’s bibliographic patent data content and don’t mind manually inspecting and deciphering the file’s tagged information as outlined above. More generally, what we want to do is examine the information content of a number (preferably a very large number) of bibliographic patent data XML files in parallel. Future posts will delve into that topic.