In October 2020, the United States Patent and Trademark Office (USPTO) released a report titled “Public Views on Artificial Intelligence and Intellectual Property Policy.”1 The report signifies the USPTO’s commitment to keeping pace with the rapidly improving Artificial Intelligence (AI) technology and to deploying AI to accelerate a wide range of the USPTO’s tools2. Indeed, the report of 2020 echoes the USPTO’s commitment to adopting long-term, comprehensive strategies for promoting and providing leadership for many AI-based tools3, previously outlined in “Remarks by Director Iancu at the Artificial Intelligence: Intellectual Property Considerations Event” in 2019.4
On March 18, 2021, Drew Hirshfeld, performing the duties of the Under Secretary of Commerce for the IP and Director of the USPTO, emphasized that the “integration of AI technologies into next generation tools offers an exciting opportunity to enhance the quality and efficiency of patent and trademark examination.”5 He pointed out to the following USPTO’s ongoing AI-related efforts:
employing AI to perform an automatic patent classification;6
employing AI to automatically route patent applications to the “right” Examiners;7
employing AI to enhance prior-art search tools;8 and
employing AI to automatically generate templates of office actions.9
In this paper, the authors present their views on how the patentees could benefit from the USPTO’s AI-based tools deployed to various USPTO’s processes.
Some of the USPTO’s efforts include employing AI to perform a patent classification. The USPTO’s auto-classification tool leverages a machine learning model to classify patent documents using the Cooperative Patent Classification (CPC) system.10 The USPTO’s patent classification tool assigns claimed subject matter in patent applications to corresponding CPC symbols.11 The tool is also configured to suggest additional CPC symbols for the identified claimed subject matter to facilitate an additional refinement of the suggested CPC symbols to be used by the USPTO’s search engines.12
The benefits that patentees may obtain from the USPTO’s use of AI to classify patents includes a reduction of the USPTO fees that the patent applicants need to pay. Any such reduction of the USPTO fees would be a result of the USPTO’s reduced costs when the patents are classified automatically rather than manually.
The USPTO could also use some of the cost savings achieved by deploying the AI-based patent classification tools to, for example, hire additional USPTO’s Examiners and thereby reduce the examination backlog. This in turn may provide an additional benefit to patent applicants, i.e., shortening the wait times for obtaining office actions during the prosecution of the patent applications.
Further benefits may include an increase in the accuracy and quality of patent classification. For example, it could make it more difficult for applicants to “game the system.” Examples of the system gaming include selecting a less-accurate title for a patent application in order to have a desired classification assigned to the application, or sanitizing a patent application by replacing, for example, business-related words by other words in attempt to avoid having the application assigned to art groups that are considered less desirable (e.g., 36xx).
The USPTO is exploring the use of the AI technology to ensure that patent applications are routed to the Examiners who are most familiar with the corresponding technologies.13 From the USPTO’s perspective, this effort is meant to reduce the costs of the patent application routing (which is presently performed manually) and to improve the efficiency of the patent application examination.14
One of the benefits that patentees will obtain from the USPTO’s AI-based tool for routing patent applications is a reduction of the application prosecution cost. Specifically, if the AI-base routing tool succeeds in routing patent applications to Examiners who are familiar with the corresponding technology, then less time will need to be spent (a) explaining inventions to Examiners, and (b) responding to rejections that are the result of a misunderstanding of the invention or the relevant prior art. Ideally, the patent applicant could save the time and the cost of at least one prosecution cycle during which the patentee would otherwise have to explain the invention to the Examiner.
Additional prosecution cost reductions may be achieved because the application is disposed of in a shorter amount of time if the application is routed to an experienced Examiner rather than if the application is routed to a less experienced Examiner. Note that this may also include a shortening of wait times for receiving first office actions since the experienced Examiners are likely to prepare the first office actions faster than the less experienced Examiners.
Moreover, if the AI-based routing of patent applications to Examiners is performed automatically, not manually as it is done currently, then the cost savings obtained by the USPTO may be passed onto patent applicants in a form of lower USPTO fees, shorter wait time (if additional Examiners are hired), and the like.
The USPTO has been considering engagements with academia and industry to help the USPTO to identify advanced patent search tools.15 For example, the USPTO has asked the industry to provide information about techniques that leverage AI for the purpose of improving the prior-art searches for examining patent applications.16
Identifying the prior art that is most relevant to an invention is an important component of the patent examination process and is critical to achieving the USPTO’s mission to grant valuable patents. However, due to the exponential growth of prior art and incredible pace of technological innovation, it is increasingly difficult to quickly discover the most relevant prior art.17 However, to meet that challenge, the USPTO has been developing an AI-based prototype search system that helps to identify relevant prior art documents and provides suggestions for additional areas to search.18 The USPTO’s system will utilize world-class patent AI models, and will be configured to automatically capture feedback data from the USPTO’s examiners to yield additional enhancements to the prior art searches over time.19
A beta version of the new AI tool was released in March 2020.20 It allows federated searches across patents, publications, non-patent literature, images, and the like. Those tools are also expected to effectively segment information, retrieve the most relevant results - not just more results, retrieve the results from a broader corpus of data, and retrieve the results quickly and efficiently.
One of the proposed solutions is a new cognitive assistant called “U,” or “Unity,” which leverages AI and machine learning to augment the existing patent tools.21 The U assistant is intended to allow Examiners to conduct, through a single-click, complex searches across various collections of documents and images, and to obtain the results organized according to various criteria and relevance levels.
Deployment of the USPTO AI-based prior-art search tool will provide patent applicants with many benefits, including obtaining the closest prior art starting from the beginning of the prosecution cycle. This approach would help patentees avoid situations of being “stuck” in a loop of overcoming rejections based on one set of weak prior art just to have the Examiner do another search that finds better (or at least different) art. In theory, this expensive cycle can continue indefinitely. By having the relevant prior art cited at the beginning of the prosecution, the patent applicants may save a great deal of prosecution time and a great deal of money. Having the most relevant prior art cited from the start of prosecution will tend to reduce the number of prosecution cycles, reduce the number of Requests for Continued Examination (RCEs), potentially eliminate the need for appeals, and the like. This is especially important since the patent applicants would like to obtain high quality patents expeditiously and at a reasonable cost.22
The USPTO is testing new AI-based tools and techniques for generating templates for office actions.23 Such tools, including robotic process automation (RPA) tools, could be configured to automatically populate the templates, and thus to save the Examiners’ time and spare the Examiners from some the tedious clerical aspects of drafting the office actions.24
As expected, the quality of such tools might be measured by the quality of the office action templates, and how much time the use of such templates saves the Examiners. In some aspects, the automated tools for generating templates for office actions may be evaluated like the tools for generating templates for replies to the office actions.
There are many automated tools available for generating templates for replies to office actions. Examples of such tools include ClaimMasterTM25, PatentOptimizerTM26, PatentBotsTM27, and the like. The tools allow generating shells of the replies, and thus allow reducing the time and cost of preparing the reply templates.
However, the difference between the reply-template-tools and the office-action-template-tools is that the reply-template-tools automatically copy the headers and rejections from the corresponding office actions, while the office-action-template-tools require the Examiners to provide the headers and rejections for the templates. Hence, the office-action-template-tools may be less automated (and therefore more burdensome), more error-prone, and more difficult to implement than reply-template-tools.
With regard to the automatization of the process for generating templates of office actions, the patentees will benefit from such tools if the tools provide accurate references to corresponding paragraphs/excerpts from the cited references on which the Examiners rely when rejecting individual limitations of the claims. While this may be difficult to implement, if implemented well the functionalities would indeed be beneficial to the patent applicants.
This article advances several benefits of the USPTO’s efforts to use AI to enhance the efficiency and accuracy of the patent examination process as the patent applications are routed through a patent examination pipeline. Patentees should benefit from the USPTO using AI to, for example, classify the patent applications because it will cause a reduction of the USPTO fees if the cost savings obtained by automating the classification are passed onto the patentees, and should benefit by having their wait times shortened if the cost savings obtained by automating the classification are used to hire additional Examiners.
Furthermore, the patentees should benefit from the USPTO’s AI-based prior-art search tools if the tools allow identifying the relevant prior art from the beginning of the prosecution of patent applications and if the tools allow avoiding tiresome prosecution loops including addressing irrelevant art from one office action to another.
Disclaimer: This article is purely a public resource of general information that is intended, but not guaranteed, to be correct and complete. It is not intended to be a source of solicitation or legal advice and is for informational and entertainment purposes only. The information is not intended to create, and receipt does not constitute, an attorney-client relationship. The laws of different jurisdictions may be implicated, and facts and circumstances can vary widely. Therefore, the reader should not rely or act upon any information in this article, but should instead seek legal counsel for individualized legal advice. For more information, please contact a firm attorney through www.hickmanbecker.com.