How AI tools will change patent practice

• English

AI-based technologies will play a significant role in patent practice and, particularly, in managing the growing demand for patents to protect increasingly complex inventions. We are already seeing some emerging new tools that support patent prosecution, searching, portfolio management and enforcement. These and other initiatives now being developed will soon become commonplace. Let us discover the role that AI technologies can play in patents and how they can support and enhance the work of patent professionals.

In the recent IP Trend Monitor survey conducted by the Dennemeyer Group and CTC Legal Media, two-thirds of respondents, and 69 percent of patent specialists, put AI ahead of all other issues in terms of its likely impact on IP work in the next five years. Tasks such as IP portfolio management, patent searching and annuity payments, in particular, were identified as likely to be affected by AI, machine learning and automation tools. A significant majority of respondents said that automation and AI would improve the efficiency of IP services. However, the survey also revealed a degree of uncertainty. Are you interested in making the most out of your patent searches? 


More than half of respondents said they were not sure whether there is currently enough useful information about artificial intelligence and its impact. Part of that uncertainty may be that, while AI-based tools can help professionals deliver results to their clients, they also raise questions about whether reforms to laws, regulations and processes will be needed. Moreover, AI tools may lead to changes in the way people work and the type of work they do. Unlike almost any other technology, therefore, AI raises fundamental questions about both IP policy and IP practice. 

AI and patent policy

WIPO recently launched a consultation on IP policy and AI and received responses from 22 member states, more than 100 organizations and over 100 individuals. BlackBerry, Robert Bosch, Ericsson, Huawei, IBM UK, Intel, Merck, Philips, Siemens and Tencent were among the companies that made submissions. The comments covered a range of topics in WIPO's Draft Issues Paper, with particular attention being paid to the implications of inventions being autonomously generated by AI. While some maintain that this possibility is not something that is pressing, the question has come into the spotlight lately with the filing of numerous patent applications for two inventions that name a machine called "Dabus" as the inventor.

The Dabus applications effectively question whether non-humans can be named as inventors on patent applications. Two decisions published so far (by the EPO and UK IPO) have accepted that the machine was indeed the inventor in these two cases but that it cannot be considered as such under existing laws. Both cases have been appealed, and these decisions, as well as those in other jurisdictions, should shed more light on this interesting and important topic.

Other issues covered in the submissions to WIPO included patentable subject matter, non-obviousness and the "person skilled in the art" test and disclosure – all of which are fundamental to the patent system. These and other relevant policy questions will be thoroughly studied by WIPO, which will produce a revised written analysis later this year. Other patent offices, such as the EPO and USPTO, are also conducting studies and hosting events looking at whether changes need to be made to patent laws and regulations. These policy questions are fundamental, and it will take time to adequately address them in a balanced way on a worldwide basis. In the meantime, AI's impact on the practical aspects of patent work is increasing.

Some examples of AI in patent practice

Patent work is highly technical and involves a significant volume of data and strict deadlines. This means it can benefit greatly from computing tools. Indeed, patent professionals have long used software for tasks such as managing annuity payments and communicating with patent offices. AI, however, offers an extra dimension. Using sophisticated algorithms, and with large sets of data, AI systems can recognize images or words (written or spoken), understand language (even in specialized contexts) and recommend or make decisions based on machine learning and previous outcomes. AI tools can assist inventors and third parties in navigating the ever-growing world of patents. With an estimated 120 million patent documents available in many different languages, and the number growing every day, AI may soon become essential to keeping the business of patents manageable and efficient.

A few ways in which AI can be used

• Patent searching: Before anyone even thinks about filing a patent, mapping the landscape and establishing the freedom to operate are essential. Yet this can be almost impossible to do manually given the number of patents and other prior art, which are growing significantly every day. AI tools can help identify the most relevant documents, based on pertinent information, and prioritize these results to save professionals time reviewing irrelevant results. This can assist in performing quality infringement analysis and provide essential information on improving the quality and validity of your applications, which is vital to minimizing legal and business risks.
• Risk management, patent analysis and valuation: Going even further, with appropriate data, AI can be used to analyze patent portfolios and evaluate their strength or weakness in particular areas. The AI tools to assist such an evaluation will grow in sophistication as more data is collected, including details regarding the claim scope, ownership and overall disclosures of the growing prior art universe.
• Translation: Thanks to advances made in language recognition, AI systems are already widely used in patent translation, something that has become more important given the exponential growth of documents in languages such as Chinese and Korean. For example, Patent Translate (developed by the EPO and Google) is based on neural machine translation and includes documents in 32 languages. Despite initial skepticism, many patent professionals now trust translations generated by this software.
• Drafting and prosecution: While communication between patent applicants and offices is now mostly electronic, the actual process of preparing and writing patent applications is still manual. However, some providers now offer AI-based patent drafting, which promises to deliver most of a completed patent application when the system is provided with certain information about the invention. Given the complexities of patent law, many people have reservations about these services, but it will be interesting to see how they develop and to what extent they can supplement the work of patent attorneys.
• Contract drafting and review: In many fields of law, AI tools based on text analysis are being used to compare documents and identify errors or changes required. These will be relevant in the patent area, particularly in agreements on assignments, employee remuneration, licensing and technology transfer.
• Dispute resolution: Patent litigation is often document-intensive. In jurisdictions with required disclosures and significant discovery obligations, e-discovery can result in thousands or even millions of files in diverse formats having to be analyzed. Tools that quickly and accurately extract relevant information from text, images, audio, databases and the like will become increasingly valuable. AI tools may also play a role in other aspects of litigation, such as predicting outcomes and quantifying damages awards. There is considerable doubt, however, as to whether computers will ever be able to replace humans as advocates or judges.

Case study: Dennemeyer Octimine

An example of the opportunities presented by AI technology is Octimine, a patent insight service provided by the Dennemeyer Group. There have long been tools available to organize and identify relevant information in patent documents, but these generally rely on classification systems and Boolean searches. Octimine, however, uses machine learning to enable users to search semantically with natural language text in a variety of formats and refine the search's scope with filters. Searches can also be made using patent publication numbers. Using sophisticated algorithms, Octimine can identify the 1,000 most similar publications from a database of more than 120 million patent documents. It can also graphically illustrate the information in these 1,000 search results, for example, to show who has filed the most applications, who is collaborating with whom, and which technologies are developing most significantly.

Because Octimine's searches are based on textual information rather than patent classification, it can identify documents beyond the limits of standard classifications. This increases the chances that relevant results will be returned. Octimine can also be customized by using synonyms that are common for a particular industry or technology to ensure that all relevant results are considered. Dr. Matthias Pötzl, the co-founder of Dennemeyer Octimine, said that AI enables devices to take over a lot of the search and some of the analysis from humans: "Such machines can find the relevant documents pretty quickly, but they cannot replace a patent searcher completely. Legal interpretation is still very hard for a machine to do."

Looking ahead, technologies such as Octimine will also transform patent monitoring. For example, alerts can be provided weekly, returning only new search results. Such powerful outcomes can be achieved with Octimine, as the neural networks are trained by the particular portfolio and the results of previous alerts. This powerful technology offers targeted information highly tailored to the user's needs, saving precious time that would be otherwise spent on sifting through irrelevant documents. Opportunities for patents AI is transforming many aspects of our daily lives, including the way we work, shop, travel and entertain ourselves. The patent practice is no exception, and change is coming quickly. "We did not see much use of machine learning in patent practice until about 2015. But awareness has really increased over the past four years and it is now getting better and better every month," says Dr. Matthias Pötzl.

Some practitioners, however, are skeptical about the impact of AI, and some understandably fear that it poses a threat to their work and skills. However, it is increasingly evident that AI promises to deliver greater efficiency, accuracy and transparency to patent work. The number of patent applications filed each year continues to rise; applications at the EPO grew 4 percent in 2019, for example, with the biggest increases coming from China and Korea. In this ever more complex world, AI-based tools will become essential to ensure that the patent system functions effectively and remains relevant to the industry's needs.

This article was first published in The Patent Lawyer Magazine, March / April 2020 issue.