EU BERT is is a language processing tool tailored for European languages and was developed through collective European efforts. EUBERT processes and understands multiple languages from the continent, addressing its specific linguistic needs.A significant aspect of EU BERT's development was the use of the EuroHPC Meluxina cluster, a core component of Europe's high-performance computing landscape. This advanced infrastructure ensured that the model was trained to be both efficient and attuned to the complexities of European languages.The model is accessible here: https://huggingface.co…

The European Parliament has deployed a EuroVoc (https://op.europa.eu/en/web/eu-vocabularies/dataset/-/resource?uri=http://publications.europa.eu/resource/dataset/eurovoc) classifier in production using the Huggingface Enterprise Hub solution. The classifier employs a BERT model to classify documents based on the EuroVoc vocabulary, which is a multilingual thesaurus of concepts used in European Union documentation.The model is accessible here: https://huggingface.co/EuropeanParliament/eurovoc_en. For more information or queries contact the solution owner via the link in the 'Contact information…

We are starting our Data Analytics effort and experimenting different tools in the Lab.Knime added to the list of tool available in the catalogue of the EP. The evaluation of Dataiku is on going.For more information or queries contact the solution owner via the link in the "Contact information" section.

We are exploring the market for OCR solutions that will help us automate some Document encoding processes. The results are available in a document that can be provided upon request. For more information or queries contact the solution owner via the link in the 'Contact information' section.

We are doing a lot of experiments with Augmented and Virtual Reality. They are all under POCs efforts.The use cases are Virtual Visits, Augmented Content, Indoor geolocalisationOur current focus is on two geolocation technologies.For more information or queries contact the solution owner via the link in the "Contact information" section.

The scope of this project is to transcribe automatically and machine translate parliamentary debates in real time and in 24 languages, including automated language recognition and an acceptable word error rate and latency. The 24 languages have been delivered by the contractor and we are now working on the final language evaluation and closing of the project.For more information or queries contact the solution owner via the link in the "Contact information" section.