・Response-Based Learning for Patent Translation
Stefan Riezler, Heidelberg University
Abstract: In response-based structured prediction, instead of a gold-standard structure, the learner is given a response to a predicted structure from which a supervision signal for structured learning is extracted. Applied to patent SMT, different types of environments such as a downstream application, a professional translator, or an SMT user, may respond to predicted translations with a ranking, a correction, or an acceptance/rejection decision, respectively. We present algorithms and experiments that show that learning from responses alleviates the supervision problem and allows a direct optimization of SMT for tasks such as cross-lingual patent prior art retrieval, or translation of technical patent documents.
|Bio: Prof. Stefan Riezler has been appointed full professor and head of the chair of Linguistic Informatics at Heidelberg University in 2010, after spending a decade in the world’s most renowned industry research labs (Xerox PARC, Google). He received his PhD in Computational Linguistics from the University of Tübingen in 1998, and then conducted post-doctoral work at Brown University in 1999. Prof. Riezler’s research focus is on machine learning and statistics applied to natural language processing problems, especially for the application areas of natural-language based web search and statistical machine translation.|
・Full-text Patent Translation at WIPO: Scalability, Quality and Usability
Bruno Pouliquen, World Intellectual Property Organization
Abstract: WIPO has 5 years’ experience in providing quality machine translation on its search engine PATENTSCOPE. Originally trained exclusively on Patent titles and abstracts, we have now experimented using descriptions and claims (full text) to train our statistical machine translation tool (called WIPO translate), based on the open source toolkit Moses. Machine translation of patent texts is now integrated in PATENTSCOPE, despite the problem of scalability (translation models trained on billions of words), quality (our automatic evaluation shows an improvement over publicly available translation sites: e.g. Google Translate) and usability (it is fully integrated in our search engine, with a translation speed less than 2 seconds per sentence).
|Bio: Bruno Pouliquen is a senior software engineer specialized in patent machine translation working at the World Intellectual Property Organization (WIPO) in Geneva since 2009. He owns a PhD in computer science (Faculty of Rennes, France, 2002) and specialized later in multilingual text mining (Joint Research Centre of the European Commission, Italy, 2001-2009), finally he works now in statistical machine translation. Focusing on building automatic machine translation tools that are used in production for many languages. He published more than 50 scientific papers in the computational linguistic domain. Bruno’s position at WIPO includes exploring Statistical Machine translation applied to the Patent domain and providing access to multilingual patent information through the PATENTSCOPE search engine.|
・Improving Translator Productivity with MT: A Patent Translation Case Study
John Tinsley, Iconic Translation Machines Ltd.
Abstract: When evaluating the suitability of MT for post-editing, there are a lot of variables to consider that could have an impact on its overall effectiveness, including: the languages in question, the documents being translated, the various workflows, and not least the individual translators themselves. Add to that the other types of evaluations – automatic measures, subjective assessments of fluency and adequacy – and we have got a lot of data on our hands. But despite all of these data points, we are ultimately trying to answer a simple question: is the MT useful and to what extent? In this talk, John will discuss a case study describing a large-scale post-editing evaluation involving more than 20 translators working on Chinese to English patent translation. He will discuss how various evaluations were carried out – from initial MT engine development to translator productivity – and discuss the implications of these findings on the real-world application of MT.
|Bio: John Tinsley is the CEO and Co-Founder of Iconic Translation Machines. He is an expert in machine translation (MT) technology, a field in which he holds a PhD from Dublin City University. The foundations of Iconic are built on methods that John pioneered over almost a decade of research and development. Prior to founding Iconic, he worked on consulting and development of MT technology for multinational clients across a variety of industries. John also acts as an expert consultant with the European Commission, providing guidance on language technology initiatives.|
・Promoting Science and Technology Exchange using Machine Translation
Toshiaki Nakazawa, Japan Science and Technology Agency
Abstract: There are plenty of useful scientific and technical documents which are written in languages other than English, and are referenced domestically. Accessing these domestic documents in other countries is very important in order to know what has been accomplished and what is needed next in the science and technology fields. However, we need to surmount the language barrier to directly access these valuable documents. One obvious way to achieve this is using machine translation systems to translate foreign documents into the users’ language. Even after the long history of developing machine translation systems among East Asian languages, there is still no practical system. We have launched a project to develop practical machine translation technology for promoting science and technology exchange. In this presentation, we introduce the background, goal and status of the project.
|Bio: Toshiaki Nakazawa received his B.S. in Information and Communication Engineering and M.S. in Information Science and Technology from the University of Tokyo in 2005 and 2007, respectively. He obtained his Ph.D. in Informatics from Kyoto University in 2010. He is currently a researcher of the Japan Science and Technology Agency (JST). His research interests center on natural language processing, particularly machine translation.|
・Initiatives of the Japan Patent Office on Machine Translation
Kei Kato, Japan Patent Office
Abstract: Recently, each Intellectual Property (IP) Office has faced common big challenges: promoting the international work sharing in the examination process and developing an environment to access foreign patent documents written in languages other than its native language. As a means of resolving such challenges, the Japan Patent Office (JPO) has actively utilized machine translation. Currently, the JPO has widely provided the foreign general users with the Japanese-English machine translated information on its patent examination results through the “One Portal Dossier” allowing them to retrieve dossier information for applications filed with the IP5 Offices (Japan, the U.S., Europe, China, and the Republic of Korea). Also, the JPO launched a new system in January 2015, utilizing the Chinese-Japanese machine translation dictionary including more than 2 million words. This system enables users to search in the Japanese language more than 12 million machine translated patent and utility model documents of the Chinese and Korean languages.
|Bio: Kei Kato has served as Deputy Director at the Patent Information Policy Planning Office, Policy Planning and Coordination Department in the Japan Patent Office since July 2015. He engages in planning policies related to the translation of patent information. He received his PhD in Engineering from the University of Tokyo. He joined the JPO as a patent examiner in 2004. In his 11 years’ career in the JPO, he has been involved in the patent examination of engine control, hybrid vehicles, switch devices, containers and packaging.|