Our mother company has been a Sponsor at European Machine Translation Summit 2019 held in Dublin on 21st, 22nd and 23rd August. During the two days prior to the main conference, several workshops were held in which new challenges in machine translation were dealt with, such as multilingualism, literature, difficulties with languages for which few resources are available and the importance of machine translation in crisis situations.
In addition, several tutorials were also held providing information on neural networks, statistics and how to choose a good set of data to perform tests. Being a Sponsor at European Machine Translation Summit has allowed Pangeanic – PangeaMT to come in contact with the best-of-class research in Europe. The main conference lasted 3 days and was divided into three areas: researchers, translators and users, in order to unite all the groups that use machine translation. Today the quality of neural machine translation is much higher and more integrated in the translation industry. However, it is necessary for translators to have an excellent command of their native language in order to add value to the content of machine translation.
Some of the most discussed topics this year have been speech translation, transfer learning, models specialized in a specific domain, the translation of languages for which little data is available and post-editing using neural machine translation models.
Presentation by Pangeanic – PangeaMT
Our chief Scientist Mercedes Garcia presented its study “incremental adaptation of neural machine translation for professional post-editors”, which is available in the proceedings, on page 219. This work was carried out together with the research group PRHLT of the Polytechnic University of Valencia in the framework of the government-sponsored “Hybrid NMT platform” project. The study consisted in incrementally updating the neuronal automatic translation models under an online learning paradigm applied in a real translation scenario. The results showed that when post-editing the results obtained from the system, the translators worked more easily and quickly. In addition, the quality of the translation and the users’ perception of this system was improved.
EU projects: NEC TM and iADAATPA
Pangeanic – PangeaMT also presented two of the projects we re heavily involved in: the NEC TM and the previous iADAATPA projects in the poster session. The aim of the NEC TM data consortium is to organize national bilingual files that have not been exploited for use as general and open machine learning data. It collects translation memories from national contract awards by Member States and helps them centralize these language resources with NEC TM’s high-performance database based on Pangeanic’s commercial database ActivaTM. iADAATPA is a platform that extends machine translation and offers machine translation providers the ability to work with customers in the European market by providing them with domain-specific engines.
The ADAPT centre also presented a poster on the evaluation of machine translation engines trained in the iADAATPA framework, showing improvements in the translation of domain-specific data provided by the institutions of the Member States of the European Union.
Sponsors at Machine Translation Summit 2019
Pangeanic – PangeaMT was one of the main sponsors of the European Machine Translation conference. We had our own table to inform attendees and speakers about our natural language processing technologies, from our Neural machine translation free panel to anonymization services.
The event concluded with the closing ceremony, in which the next MT-Summit was presented. It will take place in the United States (Seattle). In addition, it was also announced that the EAMT 2020 will take place in Lisbon and that the American chapter AAMT 2020 will be held in Florida (Orlando).