MT is not the future, but the now: Highlights from the NeTTT conference 2022 (Day 1)

This year is turning out to be a highly productive one for the translation sector, with the first NeTTT (New Trends in Translation and Technology) conference, which was held in Rhodes, Greece from July 4–6.

The conference brought together both academics and industry players with interest in translation studies, linguistics, machine translation, and other relevant domains, to share the most recent and cutting edge research and insights in the field with one another.

Right out the gate, machine translation has played a starring role, with perhaps four out of every five speeches and discussions dedicated to machine translation and MTPE (machine translation post-editing).

Of course, with such a plethora of talks available there was a lot to learn, and we’ll share some of the highlights in a series of articles. This article will tackle select proceedings from Day 1 of the NeTTT conference.

For coverage of the other days of the conference, see our articles here:

Context is key in MT: Highlights from the NeTTT conference (Day 2)
Towards better MT: Highlights from the NeTTT conference 2022 (Day 3)

Is machine translation a form of augmentation?

The conference began with a keynote speech by Sharon O’Brien on the topic “Augmented Translation: New Trend, Future Trend, or Just Trendy?”. O’Brien is a professor of translation studies at Dublin City University in Ireland, and a major name in the translation academe.

Augmentation, to put it simply, means allowing humans to overcome physical and mental limitations through technology. Translation, according to O’Brien, is already an augmented activity thanks to the use of translation management tools and the internet.

What’s interesting is that under this schema, machine translation remains an unknown factor. There is, of course, MTPE, but for the most part the potential of machine translation to augment human ability have not yet been realized to a great extent. There’s much research and innovation that needs to be done before machine translation can be considered a truly augmented activity.

More here: Is machine translation a form of “augmentation”? An interview with Dr. Sharon O’Brien

Interest in low-resource languages

One of the early panels on the first day was “Neural Machine Translation for Low-Resource Languages”, with speakers Jon Cambra and Eirini Zafeiridou. This basically focused on research into developing effective language models for languages that don’t have a substantial amount of data present on the internet.

According to Cambra and Zafeiridou, low-resource languages don’t necessarily mean smaller models. In fact, they will need more complex architectures in order to perform up to par with languages for which there is more data available.

Working with low resource languages is a research concern that has become a lot more significant lately, especially with the recent news about Meta’s pioneering open-source No Language Left Behind Project (NLLB).

Learn more about the ramifications of the NLLB project on low resource languages in our article: No Language Left Behind: Meta’s Massive Multilingual Machine Translation Ambition Pays It Forward

The perfect machine translation system?

The first day capped off with a poster session that asked the question: “What if we had the perfect translation system?” One major highlight of this panel was the discussion on the role of the translator in such a scenario.

The translator’s skillset is a crucial factor in making the best use of such a machine translation system, going beyond translation and even beyond post-editing to take on a more evolved role.

Another highlight of this discussion was the problem of authorship and copyright. Raw MT output is not protected by copyright, as it does not fulfill the criteria required to define the work done on the text as “creative”. Post-editing complicates things slightly, as it raises the question of whether the human intervention might result in the creation of an “original” work. 

The development of perfect machine translation would definitely make us rethink our notions of authorship and copyright if or when we reach that point.