Evaluation of SMT in localization to under-resourced inflected

Although Machine Translation is very popular for personal tasks, its use in localization and other business applications is still very limited. The paper presents an experiment on the evaluation of an English-Latvian SMT system integrated into SDL Trados which has been used in an actual localization assignment by a professional localization company. We show that such an integrated localization environment can increase the productivity of localization by 32.9% without a critical reduction in quality.

This paper evaluates the impact of machine translation on the software localization process and the daily work of professional translators when SMT is applied to low-resourced languages with rich morphology. Translation from English into six low-resourced languages (Czech, Estonian, 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 systems for localization purposes was evaluated 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 situation considering tag translation and its effects on the translator's productivity.

… of the Third Joint EM+/CNGL …, 2011

Although machine translation is very popular for various personal tasks, its use in business applications, including localization, is still quite limited. This paper describes the facilities of the LetsMT! platform for localization industry professionals and the results of an experiment which explored the use of the SMT system integrated into translation memory for an actual localization task. We present, LetsMT! platform from the perspective of localization, results of a user requirement analysis and our experiment of evaluating an English-Latvian SMT system integrated into SDL Trados tool. We show that such an integrated localization environment can increase the localization productivity by 32.9% without critical decrease in quality.

In cases when phrase-based statistical machine translation (SMT) is applied to languages with rather free word order and rich morphology, translated texts often are not fluent due to misused inflectional forms and wrong word order between phrases or even inside the phrase. One of possible solutions how to improve translation quality is to apply factored models. The paper presents work on English-Latvian phrase-based and factored SMT systems and, using evaluation results, demonstrates that although factored models seem more appropriate for highly inflected languages, they have rather small influence on translation results, while using phrase-model with more data better translation quality could be achieved.

2001

This paper describes a Machine Translation (MT) evaluation experiment where emphasis is placed on the quality of output and the extent to which it is geared to different users' needs. Adopting a very specific scenario, that of a multilingual international organisation, a clear distinction is made between two user classes: translators and administrators. Whereas the first group requires MT output to be accurate and of good post-editable quality in order to produce a polished translation, the second group primarily needs informative data for carrying out other, non-linguistic tasks, and therefore uses MT more as an information-gathering and gisting tool. During the experiment, MT output of three different systems is compared in order to establish which MT system best serves the organisation's multilingual communication and information needs. This is a comparative usability- and adequacy-oriented evaluation in that it attempts to help such organisations decide which system prod...

The paper describes English(Russian)-Latvian Machine Translation (MT) system which allows users with limited knowledge of the English or Russian language to understand text. The developed system has an architecture typical for transfer MT systems. It includes language identification, parsing, processing of multiword expressions (MWE), syntactic and lexical transfer, disambiguation and generation modules. Most of the system"s constituents are rulebased, however, for language identification and disambiguation statistical approach is used. One of the most complicated issues in MT is translation of MWE. The paper presents detailed analysis of the problem and provides technique for dealing with MWEs in the context of MT. Finally, the paper presents automatic evaluation results and outlines directions of further work.

ARPN journal of engineering and applied sciences, 2018

Currently, the high volume of international information exchange involves a wide range of localities. As each locality comes with its own distinctive dialect, the need for an effective means of language translation is becoming more and more apparent. Among the concerns of information professionals is the capacity of an interested party to access web information offered in an unfamiliar language. Classified under the wide field of artificial intelligence, machine translation (MT) is an approach related to natural language processing. The machine translation technique involves the use of software for the conversion of documents or verbalized information from one natural language into another. Of late, a substantial number of procedures have been proposed for the fashioning of an efficient MT system. While these procedures were observed to be capable in certain areas, they were found wanting in others. The objectives of this endeavour are to (a) conduct a thorough investigation on mach...

With a history of only around three decades, which started in the mid-1980s, localisation is a specific domain whose emergence and development benefitted from the spectacular development of multimedia and high technology. This course aims to explain the close relationship between translation and localisation, in each of the sub-fields where the contribution of trained language users is needed.

2003

The multilingual machine translation system described in the first part of this paper demonstrates that the translation memory (TM) can be used in a creative way for making the translation process more automatic (in a way which in fact does not depend on the languages used). The MT system is based upon exploitation of syntactic similarities between more or less related natural languages. It currently covers the translation from Czech to Slovak, Polish and Lithuanian. The second part of the paper also shows that one of the most popular TM based commercial systems, TRADOS, can be used not only for the translation itself, but also for a relatively fast and natural method of evaluation of the translation quality of MT systems.

2010

Translation memory (TM) plays an important role in localisation workflows and is used as an efficient and fundamental tool to carry out translation. In recent years, statistical machine translation (SMT) techniques have been rapidly developed, and the translation quality and speed have been significantly improved as well. However, when applying SMT technique to facilitate post-editing in the localisation industry, we need to adapt SMT to the TM data which is formatted with special mark-up. In this paper, we explore some issues when adapting SMT to Symantec formatted TM data. Three different methods are proposed to handle the Translation Memory eXchange (TMX) markup and a comparative study is carried out between them. Furthermore, we also compare the TMX-based SMT systems with a customised SYSTRAN system through human evaluation and automatic evaluation metrics. The experimental results conducted on the French and English language pair show that the SMT can perform well using TMX as input format either during training or at runtime.