Breaking machine translation news

Below is a brief overview of the latest news and achievements in the machine translation field. These stories were collected over the past month, but I would also highlight them among the most important news items for the last six months.

Samsung Electronics Showcases Award-Winning Machine Translation at WMT

Samsung Research is conducting research on domain-specific translation, including providing patent translation via Research’s translation service. This news story focuses on the fact that Samsung Electronics is discussing new and innovative ways to understand human language using machines and computer programs. The Language Lab at Samsung Research Global AI Center participated in a biomedical translation task, which aims to evaluate systems for translating sentences from the biomedical domain. The task covered 14 language pairs, including English, French, German and Spanish. In addition, Samsung is now acting as a new provider of machine translation, which you can test at this link: https://translate.samsung.com/. At the moment, it is quite difficult to create a new machine translation engine that can compete with the leaders of this industry. My own assumption is that Samsung’s machine translation is only suitable for certain specific topics. The article specifies biomedical translation, so I recommend that you try to use this MT in areas such as health, medicine and so on. It would also be interesting to test it on translations relating to chemistry and perhaps some other scientific fields.

Phrase Announces New Machine Translation Engine: Phrase NextMT

Phrase NextMT is a neural machine translation engine that is to be developed with a translation management system in mind, providing customers with a greater degree of customization, automation and integration. Visit https://slator.com/phrase-announce-new-machine-translation-engine-phrase-nextmt/ to find out more. Phrase is an online-only platform that operates in a browser. Phrase positions its MT engine as a universally applicable engine, so I recommend testing it for as many different language pairs and topics as possible to see how well it works. Phrase also promises that Phrase NextMT’s glossaries go beyond simple search and replace substitution to ensure that terms are correctly used and inflected. The glossary option should also be tested on as many language pairs and topics as possible. In addition, Phrase claims that the best MT engine will be applied by detecting the content type and language pairing (Autoselect). In my experience, few MT providers can provide good quality MT. For many language pairs, there are only one or two suitable providers. That is why I believe that this automatic provider selection (Autoselect) function for different language pairs and topics should be compared with human selection based on MT evaluation.

Meta Releases Open Source AI Machine Translation Model

Meta has built an artificial intelligence model — NLLB-200 — that can translate text across 200 different languages. Modelling techniques and learning will be used to improve and expand translations on Facebook, Instagram and Wikipedia. There are still many languages for which machine translation is weakly developed, or for which MT providers exist but are virtually unknown to the rest of the world. That is why the most important aspect of this story is that NLLB-200 was designed with a focus on African languages, for which it can be difficult to find sufficient data to train an AI model. “In the future, imagine visiting your favorite Facebook group, coming across a post in Igbo or Luganda, and being able to understand it in your own language with just a click of a button”. A demo using NLLB-200 to translate children’s stories from around the world is available at https://nllb.metademolab.com