Application of machine translation in localization into low-resourced languages

Application of Machine Translation in Localization into Low-Resourced Languages Raivis Skadiņš1, Mārcis Pinnis1, Andrejs Vasiļjevs1, Inguna Skadiņa1, Tomas Hudik2 Tilde1, Moravia2 {raivis.skadins;marcis.pinnis;andrejs;inguna.skadina}@tilde.lv, xhudik@gmail.com Abstract This paper evaluates the impact of ma- chine translation on the software localiza- tion process and the daily work of profes- sional translators when SMT is applied to low-resourced languages with rich mor- phology. Translation from English into six low-resourced languages (Czech, Es- tonian, Hungarian, Latvian, Lithuanian and Polish) from different language groups are examined. Quality, usability and applicability of SMT for professional translation were evaluated. The building of domain and project tailored SMT sys- tems for localization purposes was evalu- ated in two setups. The results of the first evaluation were used to improve SMT systems and MT platform. The second evaluation analysed a more complex situ- ation considering tag translation and its effects on the translator’s productivity. 1 Introduction In recent years, machine translation has received more and more interest from the localization in- dustry. To stay competitive in the market, locali- zation companies have to increase the volume of translation and decrease costs of services. For this reason, the localization industry is increas- ingly interested in combining translation memo- ries (TM) with machine translation solutions adapted for the particular domain or customer requirements. Building usable machine translation systems for less-resourced languages with complex mor- phology and syntax is difficult due to a lack of © 2014 The authors. This article is licensed under a Crea- tive Commons 3.0 licence, no derivative works, attribution, CC-BY-ND. linguistic resources, on one hand, and the com- plexity of the language, on the other hand. The benefits of the application of machine translation in localization are also recognized by developers of computer aided translation (CAT) tools. Such widely used CAT tools as SDL Tra- dos Studio, Kilgray memoQ, ESTeam Translator, Swordfish, MemSource and Wordfast besides traditional translation memory support provides integration with machine translation systems. Several cloud-based platforms offer machine translation services for the localization industry: KantanMT1, LetsMT2 and tauyou3, and others. This paper describes the methodology used for MT evaluation in localization process and results of two experiments where MT was integrated into CAT tool and used in two professional local- ization companies – Tilde and Moravia. In the first experiment we evaluated the im- pact of in-domain SMT on the productivity of translation of plain text, i.e., text without any formatting. Application of in-domain English- Latvian, English-Czech, English-Hungarian and English-Polish MT systems were evaluated by using MT plug-in to integrate them in the SDL Trados Studio translation environment. In the second experiment, we set a more com- plex scenario where translatable documents are slightly out of the domain of the SMT system, contain formatting tags, and are written in a more technical language than in the previous experi- ment. The second experiment was carried out on English-Latvian, English-Lithuanian, and Eng- lish-Estonian language pairs. In both experiments, in addition to the productivity evaluation we also performed assessment of the translation quality according to the standard internal quality …

Content truncated to 500 words