WIPO’s Terminology Database Boosts Performance Across Ten Languages

August 20, 2019

WIPO’s free terminology database, WIPO Pearl, is closing in on 180,000 terms in ten languages and now boasts a mobile-friendly interface. A new look and feel allows users to view innovative Concept Maps and AI-powered Concept Clouds directly in the term-search results hit-list and in clear full-screen mode, helping to clarify the relations among scientific and technical terms for innovators.

About WIPO Pearl

WIPO Pearl was launched in 2014 to promote accurate and consistent use of scientific and technical terms across the ten languages of publication of the Patent Cooperation Treaty (PCT), the WIPO-administered International Patent System, making it easier to search and share scientific and technical knowledge across different languages.

Scientific and technical terms can be searched in 45 different language combinations, offering the possibility to search between European and Asian languages, as well as between any of these languages and Arabic. This makes WIPO Pearl unique amongst the current offering of language resources available freely on the Web.

WIPO Pearl is developed by experienced WIPO language experts and terminologists who validate terms and assign a term reliability score. The quality of the data is high since every term is human-generated and validated by at least a second linguist, and sometimes also an expert in the technical field, prior to being deemed fit for publication.

Terms are currently derived from international patent applications filed via WIPO’s PCT system, and span 29 subject fields covering all fields of science and technology, and recently including also cutting-edge fields such as conversational interfaces (chatbots), quantum computing, green chemistry, medical robotics, and biological control. Terms come with a usage label informing the user whether the term is recommended, obsolete or standardized. Knowledge-rich contexts are provided for each term to help illustrate what it means and how it is used. Contexts are extracted from authoritative sources that are also referenced. In addition, a unique and sharable URL is now associated with the full terminology record that contains all the information on the concept and the terms.

Worldwide partners

Furthermore, an ever-growing network of partners worldwide, including universities and government bodies, now helps to populate WIPO Pearl with new concepts via term contributions made by students and terminologists, whilst subject field experts in the partner institutions check the accuracy of terms in their field of specialization and raise the term reliability score.

Concept maps

Unique to WIPO Pearl are Concept Maps that allow the user to browse domain knowledge by exploring links between concepts. These links have been researched and manually entered by terminologists so as to show concepts that are broader or narrower in scope than other concepts. Currently, over 18,000 concepts (or 85% of those present in the database) are linked to other concepts and are displayed in the Concept Maps.

Users also have the option to exploit information displayed in subsets of Concept Maps - so-called Concept Paths - to conduct combined keyword searches in PATENTSCOPE, WIPO’s patent database. This feature clusters related concepts, synonyms, abbreviations and spelling variants, thus helping with highly targeted patent document retrieval. The option to browse PATENTSCOPE and retrieve relevant patent documents is also available from the hit-list of results obtained by conducting a traditional search by term in the Linguistic Search interface.

WIPO Pearl and artificial intelligence

In addition to providing human-built and validated Concept Maps, WIPO Pearl leverages artificial intelligence to generate so-called Concept Clouds. An open-source deep learning algorithm has been trained on the full dataset of terms, contexts and manually entered concept relations so as to suggest new clusters of relations between concepts. These can then be displayed as Concept Clouds within the Concept Maps. WIPO Pearl is the first terminology database to feature such an innovation.

WIPO Pearl is also integrated with WIPO Translate, WIPO’s internally developed machine translation engine featuring innovative neural machine translation (NMT) technology, so as to offer the user machine-translated suggestions when a term is not available in one of the target languages. Whenever a concept does not appear at all yet in the terminology database, the query is redirected to PATENTSCOPE CLIR.

New features

As of May 2019, WIPO Pearl offers a new and improved look and feel, complete with mobile responsiveness. Moreover, following the recent addition of Arabic, the user interface is now available in all of the ten PCT publication languages, giving users the option to browse the application in their preferred language. Future developments of WIPO Pearl include developing interfaces allowing the dataset to be accessed from within other applications, and a fully-fledged mobile app. In due course, WIPO Pearl will also include collections from other areas of WIPO, such as trademarks, industrial designs, and terminology found in other treaties administered by WIPO, making it a complete reference resource for multilingual intellectual property terminology.