Interview with Scott Beliveau, Advanced Analytics Branch Manager, USPTO

The primary role of the United States Patent and Trademark Office (USPTO) is to grant patents for inventions as well as to register trademarks and service marks for products and services. One of the main ways in which they advance this mission is to use automation and AI in several ways to improve the operational efficiency of their patent examiners.

At the upcoming AI in Government online event in October, Scott Beliveau, USPTO Advanced Analytics Branch Leader, shares his thoughts on how the USPTO leverages data, automation and AI to advance efforts. In this interview, he identifies how a small, disjointed USPTO team created an award-winning AI / ML program that saves the USPTO tens of millions of dollars and responds to over 200 million public inquiries. per year. Analytics, Automation, and AI work together at the USPTO in a number of different use cases and examples of successful AI implementations.

In what innovative ways are you using data and artificial intelligence (AI) to benefit the USPTO?

Scott Béliveau: As a fee-based agency, data is the “liquid capital” of the USPTO – we see it as an asset to improve our internal decision-making and a way to empower entrepreneurs and innovators. The data supports our program evaluations and informs our business cases and financial analyzes. The data also enables our agency to identify cost savings, enable predictive planning, and improve how policies and programs work.

High-quality data also feeds into AI at the USPTO. Patents’ AI efforts focus primarily on natural language processing (NLP) technologies to support patent search and classification. The brands’ efforts are focused on commercial computer vision products to detect fraud. The use of these AI technologies can help us in our drive to deliver high quality and timely patent and trademark applications.

How are you leveraging automation to help you on your AI journey?

Scott Béliveau: Our path to AI has really been more of a data journey. We started by establishing a database through a shareable, “social” platform (DeveloperHub) to showcase unique ways to use our data and combine it with other datasets. People could take our data, use it, build it, and give us more information to continue the cycle. This database then allowed us to use natural language programming to extract and codify information for recognition. Today our data is used in countless areas, including inclusion in the Pile dataset, a development in the AI ​​/ NLP research community.

How do you identify which problem (s) to start with for your automation and cognitive technologies projects?

Scott Béliveau: We always start from a customer value perspective, rather than what IT can do. We then go through a series of questions such as “What do you want?” “What would you do with it if you got it?” Or “How much is it worth to you?” With answers in hand, we are focusing our efforts on securing incremental gains to support longer-term efforts.

What are some of the unique opportunities the public sector has in data and AI?

Scott Béliveau: Our agency has data covering every conceivable innovation over the past 250 years. As a public servant, I often get to meet inventors and hear their stories of how they have used data or our public AI services to start a new business or do a better job. Working in the public sector offers this unique opportunity to impact the lives of many people.

What use cases can you share where you have successfully applied AI?

Scott Béliveau: The USPTO currently has two concrete examples of AI currently in production: enriched citations and automatic classification.

The first use of AI in production at the USPTO was an effort called “Rich Quotes.” Our team used Natural Language Processing (NLP) to deconstruct responses to patent applications (known as Office Actions) and to create rich citations that made searching easier and faster for stakeholders and international partners. This approach used design thinking from a user perspective to understand stakeholder needs and the myriad of data variables required to deliver user-centric results. The NLP model has been shown to be both faster and more accurate than the previous work of dozens of experts. Using NLP has saved the agency millions of dollars in the implementation of enriched citations.

We have also deployed AI and machine learning (ML) in our patent classification efforts. Each innovation that the USPTO receives is categorized into one or more symbols from more than a few hundred thousand categories. Our current manual classification service is comparatively slow and expensive. Our new AI / ML algorithms, dubbed AutoClass, have been “trained” to classify patent and non-patent documents with classification symbols within hours, at one-tenth the cost and with similar quality. This service integrates user comments to verify and validate the accuracy of the results. AutoClass offers seamless integration into our routing and search functions with significant cost savings. This new, smarter routing system has already saved the agency and its clients time and millions of dollars.

What are the challenges for AI and ML in the public sector?

Scott Béliveau: One of the challenges we face with AI and ML in the public sector as an administrative agency is finding the right balance between explainability and transparency. Explaining the rationale for our decisions is essential to ensure trust and transparency in the intellectual property system. Transparency of training data and algorithms is of crucial importance, as any bias could have unintended negative effects on candidates. At the same time, demanding full transparency potentially opens the process to the “game” of people seeking to manipulate the process. Full transparency also potentially limits the USPTO’s ability to use private sector ML services, as many of them exploit proprietary trade secrets.

How do analytics, automation, and AI work together at the USPTO?

Scott Béliveau: Analytics, automation, and AI are all essential to our data program and our lifecycle. Our patent examiners and trademark attorneys use data every step of the way when making legal decisions about granting a patent or registering a trademark. USPTO teams perform analysis on the data captured at each step of this process to identify opportunities for improvement. We take these opportunities to improve ourselves using automation, AI / ML or non-IT activities. Finally, we use data to assess the results of these improvements; thus completing the lifelong learning cycle.

How do you deal with privacy, trust and security issues related to the use of AI?

Scott Béliveau: Carefully. Intellectual property-related industries, according to a 2016 Commerce Department study, account for 30% of employment in the United States. Failure to protect innovation (until it can be shared by law) can have disastrous consequences for a small business or for our country’s global competitiveness. Security is a major concern and most certainly determines every step of our decision to create, launch and use AI technologies.

What are you doing to develop an AI-ready workforce?

Scott Béliveau: As an agency of thousands of computer scientists and engineers who review the latest and greatest technologies every day, the USPTO has a head start in developing a workforce ready for the job. ‘IA. However, AI is a rapidly evolving field, and we’ve found that the best way to encourage AI readiness is to promote an organizational culture that continually learns and has a passion to embrace internal innovation. as much as it embraces the innovation seen in the applications received every day. Build – Measure – Learn and repeat, and if something doesn’t work, learn it and move on.

What AI technologies are you most looking forward to in the coming years?

Scott Béliveau: Collaborative intelligence. Machines are ideal for processing large amounts of data faster, freeing people up for less mundane or repetitive tasks they perform. I’m particularly interested to see how innovators are able to take advantage of advances in collaborative intelligence – not just to automate processes, but how to redesign processes to take advantage of collaborative intelligence technologies.

Scott Beliveau will be making a presentation at an upcoming AI in Government online event where he will have the opportunity to deepen these areas as part of the online virtual event.