Computing Community Consortium Blog

The goal of the Computing Community Consortium (CCC) is to catalyze the computing research community to debate longer range, more audacious research challenges; to build consensus around research visions; to evolve the most promising visions toward clearly defined initiatives; and to work with the funding organizations to move challenges and visions toward funding initiatives. The purpose of this blog is to provide a more immediate, online mechanism for dissemination of visioning concepts and community discussion/debate about them.

USPTO Seeking Text Recognition, Image Analysis Algorithms

December 20th, 2011 / in policy, research horizons / by Erwin Gianchandani

USPTO Innovation Challenge [image courtesy NASA Tournament Lab].The U.S. Patent and Trademark Office (USPTO), together with the recently created Center of Excellence for Collaborative Innovation (COECI), launched the USPTO Innovation Challenge last week, offering $50,000 in prizes for specialized algorithms that can “help bring the 7 million patents presently in the patent archive into the digital age.” In particular, the USPTO Innovation Challenge is seeking new algorithms to automatically identify and locate specific elements within patent documents, as part of a broader effort to improve the patent examination process. According to Robynn Sturm Steffen, a Senior Advisor in the White House Office of Science and Technology Policy (OSTP):

Approximately half-a-million U.S. patents are filed by inventors, entrepreneurs, and businesses each year. The Challenge is part of a broader USPTO endeavor to modernize the information technology used by the Office’s patent examiners.


Most patent examiners must look at hundreds of pages of documentation for each patent they examine, so many aspects of the new system are designed to reduce the amount of page flipping, improve readability, and allow for annotation of the documents. This includes the need for a better way for patent examiners to view and parse the detailed drawings that are often critical to determining whether an invention merits a patent [more after the jump].


A drawing from Patent No. 8033945 [image courtesy OSTP Blog].The Challenge will spur the development of new software to automatically identify key features of the technical drawings, such as the figure boundaries or the locations and text of the part labels. The drawing page [shown to the right], taken from a patent called “Bicycle Transmission System” (number 8,033,945), shows a figure and intricately labeled parts. But as is often the case with patent applications, the descriptions of those parts are buried on another page. Armed with the results of this challenge, a patent examiner would be able to see the descriptive text associated with each part right on the drawing page. For instance, part 260 will be labeled with “driving load torque on sun ratchet” and part 390 with “pawl control cable.” Having the descriptive text right next to the part number will reduce the need for the examiner to flip back and forth between the textual description and the drawings, thus making it easier to understand the proposed invention.


That clarity will eventually be helpful to everyone interested in intellectual property who uses the Office’s publicly available tools. Further, because the USPTO makes patent data publicly available for download, the results of the competition will also help improve patent analysis tools built by businesses, non-profit organizations, or anyone else who builds on the data.


Few image processing problems can be solved reliably today, but image processing experts think this particular problem is sufficiently well defined to reap the benefits of crowdsourcing, which has a strong track record of surfacing novel solutions from unexpected places. In fact, more than 900 programmers and image processing experts from around the world have pre-registered to compete in this Challenge, increasing the chance it will end up with effective algorithms. Results are expected to be selected before February 16, 2012.

Pre-registration closed on Dec. 5th, and the contest now runs through Jan. 16th. Teams of two competing in several “virtual competition rooms” have access to a training dataset of approximately 200 manually labeled patents, plus their own capabilities in text recognition, image analysis, and the construction of bounding boxes.

A total of $10,000 will be awarded to the first-place team and $5,000 to the second-place team. Another $35,000 will be distributed in the form of room prizes to the winners in the virtual competition rooms.

To learn more, including details about other challenge competitions being enabled by the COCEI, check out Sturm Steffen’s full description of the USPTO Innovation Challenge here, and view the challenge contest website here.

(Contributed by Erwin Gianchandani, CCC Director)

USPTO Seeking Text Recognition, Image Analysis Algorithms

Comments are closed.