Search for Tomorrow: Some Side Effects of Patent Office Automation

BY Andrew Chin
The United States Patent and Trademark Office’s (“Patent Office”) move to a paperless search facility and the public’s growing involvement in prior art search have recently elevated the role of search engine technology in the patent examination process.  This Article reports on an empirical study that examines how this technology has systematically changed not only how patent references are found, but also which patents are cited as prior art.
Publicly available records do not provide information identifying the method by which each of the prior art references cited by a patent was found, such as keyword search, citation tracking, or classification search.  The main methodological contribution of this Article is to identify large sets of patent citations that are likely to exhibit characteristics similar to those of citations actually found through a particular search technique.  By applying this synthetic approach to a comprehensive citation database, this study compiles a large set of patent citations that can reasonably be imputed to keyword search.
A longitudinal analysis of this imputed data set indicates that examiners became increasingly reliant on keyword full-text search in the late 1990s, as the technology became accessible from their desktop computers.  This change in examination practice appears to have had a substantive effect on the choice of patents to be cited as prior art.  Specifically, patent citations imputed to keyword search tend to be co-classified (according to the Patent Office classification system) more frequently than patent citations in general and patent citations imputed to citation tracking methods.
These findings support the concerns of some commentators about Patent Office automation and the outsourcing of prior art search.  In particular, it appears that the Patent Office classification system is not being fully utilized to improve the precision of search results.  This Article concludes with a survey of some initiatives and techniques that have recently emerged to address this problem.

The United States Patent and Trademark Office’s (“Patent Office”) move to a paperless search facility and the public’s growing involvement in prior art search have recently elevated the role of search engine technology in the patent examination process.  This Article reports on an empirical study that examines how this technology has systematically changed not only how patent references are found, but also which patents are cited as prior art.

Publicly available records do not provide information identifying the method by which each of the prior art references cited by a patent was found, such as keyword search, citation tracking, or classification search.  The main methodological contribution of this Article is to identify large sets of patent citations that are likely to exhibit characteristics similar to those of citations actually found through a particular search technique.  By applying this synthetic approach to a comprehensive citation database, this study compiles a large set of patent citations that can reasonably be imputed to keyword search.

A longitudinal analysis of this imputed data set indicates that examiners became increasingly reliant on keyword full-text search in the late 1990s, as the technology became accessible from their desktop computers.  This change in examination practice appears to have had a substantive effect on the choice of patents to be cited as prior art.  Specifically, patent citations imputed to keyword search tend to be co-classified (according to the Patent Office classification system) more frequently than patent citations in general and patent citations imputed to citation tracking methods.

These findings support the concerns of some commentators about Patent Office automation and the outsourcing of prior art search.  In particular, it appears that the Patent Office classification system is not being fully utilized to improve the precision of search results.  This Article concludes with a survey of some initiatives and techniques that have recently emerged to address this problem.

 

DOWNLOAD PDF | 87 N.C. L. Rev.1617 (2009)