By: Anna Xu Anton Rizor


With intellectual property becoming increasingly important in the digital age, the fact that many small and medium enterprises (SMEs) in the EU do not fully exercise their IP rights, if at all, offers a great question to those in the legal field on how we can assist and improve these parties’ access to justice.

So, it is along these lines the Conflict Analytics Lab has begun experimenting with solutions utilizing AI. In particular, the Trademark project used EUIPO data, and has begun to explore how machine learning and deep-learning models for text processing and image recognition can be used to develop suitable IP self-help and Alternative Dispute Resolution (ADR) systems for SMEs.


Machine learning and deep learning fall under the generalized category “artificial intelligence” and have progressed significantly in recent years. In fact, they have reached a point at which judgments, including those of European Union courts, can be used not only for the lawyer to understand the application of legal principles, but also for computers to apply that reasoning to create a computerized prediction for new fact patterns.

Statistical machine learning has shown great promise for the development of predictive tools capable of evaluating the odds of winning cases or estimating damages. The application of AI to law has the potential to shed new light on how legal decisions are made, illuminating the evolution of case law and the consistency (and predictability) of judicial decisions. And most importantly for our purposes, providing access to efficient AI systems could create invaluable tools for improving access to justice, currently a prominent issue in Europe and North America.

In this regard, the sub-field of trademark law and Likelihood of Confusion seems particularly well suited to the application of data science insofar as they are fact-driven areas that relies on set factors on an outline defined by a European Union trade mark (EUTM) regulation.

While machine learning and algorithms cannot replace good legal, interpretation, and negotiation skills, data science can inform an SME’s decision to obtain legal counsel, especially when it comes to finding similar trademarks or determining the Likelihood of Confusion between trademarks. This could potentially save these small businesses tons of legal fees they often cannot afford, while still allowing them to access and exercise their trademark rights.


So far, the Trademark project has made great strides in building a machine learning tool, which has already shown promising results.

To build an algorithm and create a deep learning program, data input is essential and the core of the program itself. Therefore, since September, our team has been focused extensively on annotating and inputting EUIPO case law into our database—extracting information such as the visual, phonetic and conceptual similarities between disputed trademarks, amongst other relevant factors which judges have used to determine their Likelihood of Confusion.

Though far from finished, and with a ways to go yet in our research, the preliminary results we have obtained already seem very promising, and the team working under the Trademark project are optimistic about the future of not only AI in the sub-field of IP, but its potential in the legal field as a whole.