How AI Makes Patent Search Better

Engineers can search 136 million prior art documents.

IP.com has sponsored this post.

The patent search is an integral part of the patent and intellectual property process. Because patents protect original inventions, methods, chemical compositions and machines, it is essential for patent office examiners to know if a patent application describes an original invention. Therefore, they must search the prior art to compare a patent application to existing patents around the world. Because patent applications and patent litigation can be costly, inventors and companies must also search the prior art.

Searching patent literature is not a trivial pursuit. Traditionally, the public has relied on patent libraries. The United States Patent and Trademark Office (USPTO) patent library does include a digital database; however, searching for information on government websites is not easy. The USPTO website has a seven-step process for conducting a “preliminary search of U.S. patents and published patent applications.” The process includes retrieving full-text U.S. patents, full-text patent applications and additional patent publications, then reviewing each one in-depth to “examine the selected published patent applications closely, paying close attention to the additional drawings pages, the specifications and especially the claims.” Presumably, at some point during this process, there is a risk of dying of boredom.

There are other avenues for patent searching. Conventional search engines such as Google index patents, and also use natural language search algorithms to surface relevant results for words linked to the words used as search terms. However, because these search engines are not dedicated to patent literature and it may not be clear what patent databases they are indexing, it is difficult to ensure the search is exhaustive and nothing has been missed.

That’s where IP.com comes in. IP.com has trained its AI semantic search system using unsupervised deep learning with a training data set of millions of real patent literature documents from patent libraries around the world. With the IP.com search engine InnovationQ, engineers and inventors have an easy way to search and review patent literature to speed the task of understanding patentability of an idea before pursuing protection, whether that is through a patent or by publishing.

In addition to InnovationQ, IP.com also delivers a platform called Ideas Plus, designed to serve inventors throughout the innovation process, from ideation to commercialization.

Engineers no longer have to go through documents manually; instead, they can simply pick the documents and do analysis. With the semantic search engine, it is easier to ingest the documents.

“Semantic search understands the meaning of the words, rather than literal matching of the words. It always helps to give context to improve quality of the retrieval,” said Joshua Volpe from IP.com. When doing a search on IP.com, adding a few supporting details, such as a little about how you want to use the idea or what it does, the AI can better represent what you want and dramatically improve your search experience. “We always think about talking to it like we’re talking to a person,” Volpe added. “The computer can do really good semantic analysis, but it can’t read your mind.”

By interpreting context and intention between the search terms, the search can focus on specific areas of interest and narrow significantly. For example, in InnovationQ a simple search of “Autonomous Vehicles” returns 1,500 results, and the system ranks most of these results with 5-star relevancy. The results range from sensor systems to algorithms to control systems.

But for comparison, a search of “Autonomous vehicle with lidar sensor using a single photon avalanche diode” sees the system return results which are much more narrowly focused and relevant to your needs. InnovationQ is also able to rank relevancy more usefully, calling a few hundred results “four star” and a few thousand “three star” and so on.

In this specific example, the search can even return results with “SPAD” in the title, as “SPAD” stands for “single photon avalanche diode.” Because semantic search understands the meaning of words, it is able to make this connection. The system also has a wide range of filters and tools that can help make searches more robust as you go along. This is part of how InnovationQ can support narrow searches to home in on a specific area, and also landscape analysis where a broad search is preferred.

Better Patent Search Saves Time

With so much patent literature out there, including the 136 million documents indexed by InnovationQ, time is short for an engineer or inventor to read, search and begin to understand the state of the art. “Using a tool like InnovationQ, you may not need to read and understand every document, but you’ll get a better understanding of the relevant prior art,” said Volpe.

Given this staggering number of prior and existing patents, discouraged inventors may wonder:  is there anything left to invent? Then again, it is important to remember that in research and development labs across the world, there are departments and teams whose only job is to think of new ideas and inventions, and new technologies to open up new areas of development. As opposed to closing off, patents should theoretically increase in volume as technology broadens and deepens. As such, effective patent search capability will only become more important, especially in new and emerging fields. Autonomous driving is a perfect example of this.

Who is AI-Enabled Patent Search For?

In this article, we have focused strongly on the inventor mindset, and the capability that a robust patent search tool—and the associated patent quality analysis tools at IP.com—can add for inventors and engineers. But for most companies, intellectual property protection is of concern to the legal or intellectual property department, as well. While it is Engineering’s job to come up with the patentable ideas, develop and build products, develop software or improve processes and machines, it is the job of lawyers to allocate resources to protect those ideas—and in many cases, to determine which ideas are worth patenting.

In many ways, InnovationQ is just as valuable to a lawyer examining a potential company patent before filing as it is to the engineer. Fortunately, IP.com is developed with legal department users in mind as well. Along with a stable of other tools, the USPTO has access to IP.com and InnovationQ Plus for the purpose of searching the patent literature during the examination process.

Whether you are working in an R&D lab for a cutting-edge industry or working as a manufacturing engineer incrementally improving methods and machines, patent protection can be a valuable asset for your company both strategically and financially. With limited time to tackle the vast body of prior art, it can be a challenge to adequately assess the originality and value of new ideas. With tools like those provided by IP.com, engineers, inventors and the legal department can quickly and easily assess the state of the art, identify gaps, pinpoint areas of improvement and ultimately craft better disclosures that maximize resources put toward patent protection. In any competitive engineering industry, these tools can be a significant asset.

To learn more, visit IP.com.