Abstract:The Japan Patent Office (JPO) utilizes the Machine Translation Platform
(MTP) for translating Japan’s patent publications and examination documents into English. The MTP is also used to translate Chinese and Korean patent publications into Japanese.
Neural Machine Translation (NMT) is the main engine of the MTP.
Statistical Machine Translation (SMT) and Rules Based Machine Translation (RBMT) are also used partially.
The English and Japanese translations prepared by the MTP are provided free of charge to the public through the Japan Platform for Patent Information (J-PlatPat) and the One Portal Dossier (OPD). Japan’s and foreign examiners can refer to the translations, too. The MTP contributes to disseminate Japan’s patent information to the abroad, and to research worldwide patent publications.
In this presentation, the JPO's initiatives in translating patent information, focusing on the MTP are introduced.
Abstract:Translators are communicators who consider the target reader and the culture of the target reader community. Translators are trained to familiarise themselves with the norms of the target language through explanation, clarification, or reproduction. However, translators may have a tendency to translate literally or repeat frequently used vocabulary and grammar to “play it safe” and avoid errors; thus, research on translation features is valuable. This talk will outline the results of two diachronic, corpus-based projects. The data consists of Chinese patent claims and abstracts, English patent abstracts, and English translations of Chinese patent abstracts. With quantitative and qualitative analyses of these corpora, the linguistic features of the Chinese patent terminologies, the development, and variations of patent terminologies can be identified. The translation-specific features and patterns of patent translation will also be compared with accumulated human translations of patent texts over a long-term period. Findings will present (1) features of English translations, namely of patent abstracts, and the frequency and patterns of simplification, explicitation, normalisation, and levelling-out; (2) specific patterns of translated patent texts and comparable nontranslated patent texts from different periods; and (3) linguistic features of English translations of patent abstracts of the same period. The authentic material collected in the patent corpora provides representative research data, which in turn, increases our understanding of the text features of patents.