USPTO set to award $70m contract to improve AI-driven patent search
Office intends to sign five-year contract with Accenture Federal Services | Deal includes maintenance of patent search AI capabilities and enhancements to system.
The US Patent and Trademark Office (USPTO) is set to award $70 million to a contractor for maintaining and enhancing its AI-powered patent search.
Last week (February 15), the office announced its intention to grant the five-year contract to Accenture Federal Services (AFS), which was “the only known qualified source” for the job.
Under the contract, expected to be awarded on April 1, AFS will provide a “full system development effort” to maintain the operation of the USPTO’s patent search AI capabilities and provide new enhancements to the component.
AFS represents government agencies in the US and is based in Arlington, Virginia. It is a subsidiary of Fortune Global 500 company Accenture, a private consulting firm headquartered in Dublin.
Deploying AI for prior art search
Over 9,000 examiners and supervisors use the office’s Patents End-to-End (PE2E) Search tool, which has AI functionality built in to improve prior art search.
PE2E is a web-based system that provides examiners with a set of tools to use in the patent examination process. PE2E Search sits within this system, and contains both US and foreign published documents.
The USPTO has been incorporating AI capabilities into its search system for some years, introducing the AI-based ‘More Like This Document’ feature in 2021 and the ‘Similarity Search’ in 2022.
The ‘Similarity Search’ feature allows examiners to input certain information including Cooperative Patent Classification (CPC) symbols, text snippets and words, and uses AI models to generate a list of domestic and foreign patent documents that are similar.
In August last year, the USPTO said it was seeking “to broaden its understanding of the commercial marketplace in order to improve prior art search through the deployment of artificial intelligence”.
Through a request for information, it invited potential partners to demonstrate their solutions to meet its requirements for patent search AI, including machine learning operations.
‘Responsible’ generative AI
In May last year, USPTO director Kathi Vidal said the office would “love” to use generative AI but that it had to be “responsible” before being incorporated.
This could include in the USPTO’s trademark work, she added, for example in image searches.
Speaking at the International Trademark Association’s annual meeting in Singapore, she also expressed concern over data usage and AI.
“If you’re making filings at the USPTO and you expect that information to be kept confidential, you probably don’t want us feeding it into an AI system to become part of a broader data set,” said Vidal.
Earlier this month, the USPTO issued guidance to judicial boards on the use of AI in legal proceedings.
This clarified existing regulations and outlined their applications when AI is used in the preparation of submissions for the Patent Trial and Appeal Board and Trademark Trial and Appeal Board.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
