PRESEMT

Experiments on Domain Adaptation for Patent Machine Translation in the PLuTO project Alexandru Ceau�u, John Tinsley, Jian Zhang, Andy Way Centre for Next Generation Localisation School of Computing Dublin City University, Ireland {aceausu;jtinsley;zhangj;away}@computing.dcu.ie Abstract The PLUTO 1 project (Patent Language Translations Online) aims to provide a rapid solution for the online retrieval and translation of patent documents through the integration of a number of existing state-of-the-art components provided by the project partners. The paper presents some of the experiments on patent do- main adaptation of the Machine Transla- tion (MT) systems used in the PLuTO project. The experiments use the Interna- tional Patent Classification for domain adaptation and are focused on the Eng- lish�French language pair. 1 Introduction The European Commission has supported human language technologies, in particular Machine Translation (MT), for over 40 years. This has led to a number of pioneering developments in these areas. This support has been particularly concert- ed in the past decade due to changes in the com- mercial landscape in Europe, where research in- dicates that consumers feel constrained to buying only in their own language due to issues with language barriers. A core aspect of the Commission's commit- ment to language diversification is the provision of multilingual access to intellectual property information, namely patents. This will afford inventors in Europe better access to technical information on patents in their native language and foster innovation and growth. Central to 1 http://www.pluto-patenttranslation.eu © 2011 European Association for Machine Translation. such a provision is the availability of high- quality search and translation technologies capa- ble of dealing with the volume and language di- versity of large collections of patent data. MT software must also be adapted to handle the spe- cific language found in patent documents. To this end, the European Commission has part-funded the PLuTO (Patent Language Translations Online) project to develop a framework in which their users can exploit state-of-the-art MT to translate patent documents. As well as supporting the translation needs of the Commission, PLuTO serves a more general purpose when it comes to intellectual proper- ty-related activities. There are considerable trans- lation requirements throughout the end-to-end patent application process. The necessary quality and quantity of translations varies greatly de- pending on the stage in the process. For example, at the patentability/prior-art searching stage, doz- ens of documents need to be translated but the quality does not need to be perfect; on the contra- ry, when establishing freedom to operate, a small number of documents must be precisely translat- ed as there are legal implications involved. At present, there are a limited number of tools that can carry out such translations adequately; at least not for what might be deemed an economi- cal price. Small- and Medium-size Enterprises and individual inventors can encounter difficul- ties when entering a new market due to the high costs related to translation. Often, making such a leap constitutes a large risk for these entities. Additionally, local patent agencies � who typi- cally provide expert patent translation …

Content truncated to 500 words