The paper describes the 3 NMT models submitted by the eTranslation team to the WMT 2021 news translation shared task. We developed systems in language pairs that are actively used in the European Commission’s eTranslation service. In the WMT news task, recent years have seen a steady increase in the need for computational resources to train deep and complex architectures to produce competitive systems. We took a different approach and explored alternative strategies focusing on data selection and filtering to improve the performance of baseline systems. In the domain constrained task for the French–German language pair our approach resulted in the best system by a significant margin in BLEU. For the other two systems (English–German and English-Czech) we tried to build competitive models using standard best practices.

The paper describes the submissions of the eTranslation team to the WMT 2020 news translation shared task. Leveraging the experience from the team’s participation last year we developed systems for 5 language pairs with various strategies. Compared to last year, for some language pairs we dedicated a lot more resources to training, and tried to follow standard best practices to build competitive systems which can achieve good results in the rankings. By using deep and complex architectures we sacrificed direct re-usability of our systems in production environments but evaluation showed that this approach could result in better models that significantly outperform baseline architectures. We submitted two systems to the zero shot robustness task. These submissions are described briefly in this paper as well.

This paper describes the submissions of the eTranslation team to the WMT 2019 news translation shared task. The systems have been developed with the aim of identifying and following rather than establishing best practices, under the constraints imposed by a low resource training and decoding environment normally used for our production systems. Thus most of the findings and results are transferable to systems used in the eTranslation service. Evaluations suggest that this approach is able to produce decent models with good performance and speed without the overhead of using prohibitively deep and complex architectures.

MultiUN is a multilingual parallel corpus extracted from the official documents of the United Nations. It is available in the six official languages of the UN and a small portion of it is also available in German. This paper presents a major update on the first public version of the corpus released in 2010. This version 2 consists of over 513,091 documents, including more than 9% of new documents retrieved from the United Nations official document system. We applied several modifications to the corpus preparation method. In this paper, we describe the methods we used for processing the UN documents and aligning the sentences. The most significant improvement compared to the previous release is the newly added multilingual sentence alignment information. The alignment information is encoded together with the text in XML instead of additional files. Our representation of the sentence alignment allows quick construction of aligned texts parallel in arbitrary number of languages, which is essential for building machine translation systems.

The European Commission's (EC) Directorate General for Translation, together with the EC's Joint Research Centre, is making available a large translation memory (TM; i.e. sentences and their professionally produced translations) covering twenty-two official European Union (EU) languages and their 231 language pairs. Such a resource is typically used by translation professionals in combination with TM software to improve speed and consistency of their translations. However, this resource has also many uses for translation studies and for language technology applications, including Statistical Machine Translation (SMT), terminology extraction, Named Entity Recognition (NER), multilingual classification and clustering, and many more. In this reference paper for DGT-TM, we introduce this new resource, provide statistics regarding its size, and explain how it was produced and how to use it.

This paper deals with translation of English documents to Oromo using statistical methods. Whereas English is the lingua franca of online information, Oromo, despite its relative wide distribution within Ethiopia and neighbouring countries like Kenya and Somalia, is one of the most resource scarce languages. The paper has two main goals: one is to test how far we can go with the available limited parallel corpus for the English ― Oromo language pair and the applicability of existing Statistical Machine Translation (SMT) systems on this language pair. The second goal is to analyze the output of the system with the objective of identifying the challenges that need to be tackled. Since the language is resource scarce as mentioned above, we cannot get as many parallel documents as we want for the experiment. However, using a limited corpus of 20,000 bilingual sentences and 163,000 monolingual sentences, translation accuracy in terms of BLEU Score of 17.74% was achieved.

This paper describes the acquisition, preparation and properties of a corpus extracted from the official documents of the United Nations (UN). This corpus is available in all 6 official languages of the UN, consisting of around 300 million words per language. We describe the methods we used for crawling, document formatting, and sentence alignment. This corpus also includes a common test set for machine translation. We present the results of a French-Chinese machine translation experiment performed on this corpus.

Recent developments on hybrid systems that combine rule-based machine translation (RBMT) systems with statistical machine translation (SMT) generally neglect the fact that RBMT systems tend to produce more syntactically well-formed translations than data-driven systems. This paper proposes a method that alleviates this issue by preserving more useful structures produced by RBMT systems and utilizing them in a SMT system that operates on hierarchical structures instead of flat phrases alone. For our experiments, we use Joshua as the decoder. It is the first attempt towards a tighter integration of MT systems from different paradigms that both support hierarchical analysis. Preliminary results show consistent improvements over the previous approach.

In current phrase-based Statistical Machine Translation systems, more training data is generally better than less. However, a larger data set eventually introduces a larger model that enlarges the search space for the decoder, and consequently requires more time and more resources to translate. This paper describes an attempt to reduce the model size by filtering out the less probable entries based on testing correlation using additional training data in an intermediate third language. The central idea behind the approach is triangulation, the process of incorporating multilingual knowledge in a single system, which eventually utilizes parallel corpora available in more than two languages. We conducted experiments using Europarl corpus to evaluate our approach. The reduction of the model size can be up to 70% while the translation quality is being preserved.

We present a large parallel corpus of texts published by the United Nations Organization, which we exploit for the creation ofphrase-based statistical machine translation (SMT) systems for new language pairs. We present a setup where phrase tables for these language pairs are used for translation between languages for which parallel corpora of sufficient size are so far not available. We give some preliminary results for this novel application of SMT and discuss further refinements.