Building Machine Translation Tools for Patent Language: A Data Generation Strategy at the European Patent Office
2934Authors
nan
Abstract
AbstractThe European Patent Office (EPO) is an international organisation responsible for granting patents and promoting global cooperation in the intellectual property world. With three official languages (English, German, French) and a need to constantly access and manipulate information in multiple languages, machine translation is essential for the EPO. Over the last years we have developed internal machine translation engines, specifically for the translation of patent language. This article presents our data generation strategy: it describes our approach to the generation of parallel corpora of documents, training datasets of aligned sentences, and respective evaluation datasets. Details on the challenges and technical implementation are presented, as well as statistics of the training dataset generation process.
Keywords
Full Content
© 2023 The authors. This article is licensed under a Creative Commons 4.0 licence, no derivative works, attribution, CC-BY-ND. Building Machine Translation Tools for Patent Language: A Data Generation Strategy at the European Patent Office Matthias Wirth1*, Volker D. Hähnke1, Franco Mascia1, Arnaud Wéry1, Konrad Vowinckel1, Marco del Rey2, Raúl Mohedano del Pozo2, Pau Montes2, Alexander Klenner-Bajaja1 1 European Patent Office Patentlaan 2 2288EE Rijswijk, Netherlands * mwirth@epo.org 2 European Patent Office Bob-van-Benthem-Platz 1 80496 München, Germany Abstract The European Patent Office (EPO) is an international organisation responsible for granting patents and promoting global co- operation in the intellectual property world. With three official languages (Eng- lish, German, French) and a need to con- stantly access and manipulate information in multiple languages, machine translation is essential for the EPO. Over the last years we have developed internal machine translation engines, specifically for the translation of patent language. This article presents our data generation strategy: it describes our approach to the generation of parallel corpora of documents, training datasets of aligned sentences, and respec- tive evaluation datasets. Details on the challenges and technical implementation are presented, as well as statistics of the training dataset generation process. 1 Introduction and Background The mission of the European Patent Office (EPO) is “to deliver high-quality patents and efficient services that foster innovation, competitiveness and economic growth.” (European Patent Office, 2023a). The EPO is an international organisation with English, French and German as official languages (Article 14(1) of the European Patent Convention, (EPC)) and, as a global player, it develops and promotes international cooperation at worldwide level with organisations both inside a and outside of the patent system (European Patent Office, 2023b). Both its role as a patent granting authority and being a global stakeholder in the intellectual property world requires constant access, exchange and manipulation of information in a myriad of different languages, making machine translation an indispensable tool. Not surprisingly, the most significant part of the machine translations performed concerns the translation of patent documents. A patent is a technical and legal document that gives inventors for a time-limited period the right to prevent others from creating, using, or selling their invention without their permission in the countries for which the patent has been granted. The basic legal requirements for a patent to be granted are, that the claimed invention is considered to be new and that it involves an inventive step in view of the state-of-the-art. According to the EPC, “the state-of-the-art shall be held to comprise everything made available to the public by means of a written or oral description, by use, or in any other way, before the date of filing of the European patent application” (Article 54(2) of the EPC). As will be appreciated, this definition imposes no restriction on language, i.e. in order to assess the basic requirements of patentability, examiners need to be able to access information in any possible language. However, this is not the only use case for machine translation of patent documents. In the …
Content truncated to 500 words
Labeled Text
AbstractThe European Patent Office (EPO) is an international organisation responsible for granting patents and promoting global cooperation in the intellectual property world. With three official languages (English, German, French) and a need to constantly access and manipulate information in multiple languages, machine translation is essential for the EPO. Over the last years we have developed internal machine translation engines, specifically for the translation of patent language. This article presents our data generation strategy: it describes our approach to the generation of parallel corpora of documents, training datasets of aligned sentences, and respective evaluation datasets. Details on the challenges and technical implementation are presented, as well as statistics of the training dataset generation process.
Publication Details
- Paper ID:
- 2934
- Year:
- 2023
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan