WIPO Launches State-of-the-Art Artificial Intelligence-Based Image Search Tool for Brands

Geneva, April 1, 2019
PR/2019/831

The World Intellectual Property Organization (WIPO) has launched a new artificial intelligence (AI)-powered image search technology that makes it faster and easier to establish the distinctiveness of a trademark in a target market.

Earlier-generation image search tools primarily determine trademark image similarity by identifying shapes and colors in marks. WIPO’s new AI-based technology improves on this technology by using deep machine learning to identify combinations of concepts – such as an apple, an eagle, a tree, a crown, a car, a star – within an image to find similar marks that have previously been registered.

The new technology results in a narrower and more precise group of potentially similar marks, facilitating greater certainty in strategic planning for brand expansion into new markets. With fewer results to scrutinize, this also translates into labor-cost savings for trademark examiners, attorneys and paralegals, industry practitioners and researchers.

“In the field of trademarks, our state-of-the-art AI technology is a major improvement that will create greater certainty for the development of new image marks and greater ease for monitoring potentially misleading or conflicting new registrations,” said WIPO Director General Francis Gurry. “This kind of enhanced business intelligence is invaluable in a globalized economy in which the volume of economic agents seeking brand protection is expanding rapidly.” 

WIPO’s new AI search technology leverages deep neural networks and figurative elements classification data from the Madrid System for the International Registration of Marks and from large trademark offices.

All users can access the AI search technology for free through WIPO’s Global Brand Database, where it has been fully integrated into the database search engine.

Search coverage

The new search functionality covers the national collections of 45 trademark offices already participating in the project - even if they have not been using a classification system for figurative elements. This represents a total number of almost 38 million trademarks to date. WIPO periodically adds new collections from around the world to the database.

“The increasing demand for IP rights across the globe is overwhelming current-generation systems, which is why WIPO is leading on the development of artificial intelligence-based tools that improve the global IP system,” said Mr. Gurry, adding: “A larger data pool means better AI results, so I encourage trademark offices whose collections are not included in the Global Brand Database to consider adding them as soon as they can.” 

The AI image similarity algorithm allows users to combine it with any other search criteria, for example restricting the results list to a given set of jurisdictions or to one or several parts of the Nice Classification – an international classification of goods and services applied for the registration of marks.

Users submitting a complex or composite image may use an in-built editing tool for close cropping of a searched region of interest in the image, further simplifying the searched image for more relevant results.

Background on AI at WIPO

WIPO continues to explore new technologies to ensure the highest efficiency and quality of the global IP system. A new AI-focused research and development team has developed several new tools, including the AI-powered trademark search function and WIPO Translate, a neural machine translation tool for patent texts.

In January, WIPO released the first installment in its new "Technology Trends" research series – a report that analyzes data in patent applications and scientific publications to glean the latest trends in the field.  It defined and measured innovations in AI, uncovering more than 340,000 AI-related patent applications and 1.6 million scientific papers published since AI first emerged in the 1950s, with the majority of all AI-related patent filings published since 2013.