Like many insight-driven organizations, the United States Patent and Trademark Office (USPTO) is leveraging data analytics and technologies, such as AI and machine learning (ML), to improve operational efficiency and performance, and to improve systems and processes. We are improving quality.
AI and ML algorithms are critical to agency efforts, but the guiding principle for agencies is to take a people-first approach when developing and using these technologies to improve and scale initiatives. . AI and ML tools can help enhance the work of human professionals and enhance their ingenuity in their work, but currently cannot match the nuances and reasoning capabilities of the human mind. said USPTO CIO Jamie Holcombe.
To complement technology, government agencies rely on input from thousands of experienced workers, both passively and actively collected, to train and refine AI-driven models, enabling technology to meet expectations. ensure that it delivers the expected results. The institution has awarded over 11 million patents since its inception and employs over 12,000 of his people, including engineers, lawyers, analysts and computer specialists. A continuous stream of feedback from our frontline patent examiners to improve our AI/ML models to drive new product development and to support his work in two key areas: patent search and classification. Also used for
Conducting a comprehensive patent search can be difficult given the sheer volume of data and possible sources of “prior art,” notes Holcombe. To meet this challenge, our technical team is introducing an AI component to our new patent search tool to help examiners find the most relevant sources of information they need when probing an application. This is important because each of the more than 600,000 applications the USPTO receives on average each year contains approximately 20 pages of text and figures. His IT organization at a government agency also developed and deployed a classification tool that identifies and matches classification symbols associated with inventions from over 250,000 possible categories.
In both cases, the models are developed and continuously enhanced with input from human experts. Experts determine whether something is truly new or new and apply law, facts, and expertise to make decisions.
Exploring Human Paths in Information Flow
Having a constant stream of feedback from examiner experts and others can be an asset, but the USPTO has the innovation and global expertise to help solve key challenges and scale AI. is not the only way we are taking to identify new channels for Earlier this year, the agency turned to his AI research community and Google. Kaggleis the preeminent technical and social platform used by data scientists and others to exchange thoughts and ideas.it launched globally global coding contest In March, it offered a $25,000 prize, inviting AI researchers and data scientists to write code to assess the semantic similarity of phrases.
The contest attracted over 42,900 entries and closed on June 30th. Over 1,800 global teams collaborated and leveraged publicly available resources. patent data sourceThe aim of the competition was to use AI to make a difference in understanding patent language for the agency and patent community, Holcombe explains. “As a result, not only will his algorithms for patent searches improve, but the winning model will become part of the public domain,” he says.
The USPTO also made use of other public information resources such as: goldenis a free “Wiki-style” AI/ML-driven platform launched in 2019 that scours the web to match topics and relevant available data and organizes them into information flows. AI algorithms working behind the scenes keep adding relevant data as it becomes available. Anyone can search for information about companies, patents, venture capital, and other funding sources.
AI/Human Alliance A, B, C
Much has been written about the convergence of technologies, but taking a “human-centric” approach to AI and ML development can be difficult given the diverse and complex differences in human nature. To keep the effort on track, the USPTO has created a guide to progress from pilot to prototype to production under Holcombe’s direction. Here’s the basic alphabetical order of that guide:
A is for alignment. The USPTO’s IT chief says there must be a strong connection between business and IT staff. “The best cross-functional teams are made up of technical staff working alongside business people in an agile environment that encourages planning, execution, checking, and coordination.” Agile and/or “DevSecOps” practices relies on rapid movement, transparency, and product thinking. To maximize progress, leaders engage with teams and stakeholders early and often.
B is the business value. Start with a business case that has clear value to your core strategic operations. Such use cases should address challenges where AI and ML can logically help. “as 100% commission-funded agency, Our team approaches technical challenges through a rigorous business and ROI lens,” notes Holcombe.
C is for customers (and employees). AI/ML solutions are designed to enhance, not replace, examiners and other subject matter experts. So the new technical team tests and refines the concept with internal customers before, during and after launch. Examiners using the product help drive AI innovation. Some examiners “in detail” work together in the CIO’s office to provide key information. “Because we weave customers into the process early on, we get strong feedback that helps drive adoption,” he said. “Customers also keep us honest in deploying AI that holds them accountable to government professionals and the public we serve.”
