Towards better MT: Highlights from the NeTTT conference 2022 (Day 3)

This is the third and final installment in a series covering our attendance at the first New Trends in Translation and Technology conference held at Rhodes, Greece.

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

MT is not the future, but the now: Highlights from the NeTTT conference 2022 (Day 1)
Context is key in MT: Highlights from the NeTTT conference 2022 (Day 2)

Again, 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.

For machine translation, the third day’s panels brought more focus on the topic of MT evaluation, and on inclusiveness in MT development. Here are some of the highlights from our attendance.

Machine translation for the European Union

The day started off with a keynote titled “The power of people and technology in DGT’s translation ecosystem”, presented by Merit-Ene Ilja, Translation Director at the DGT of the European Commission.

DGT refers to the Directorate-General for Translation, the Commission’s department responsible for translating its texts among the EU’s 24 official languages.

Ilja’s discussion of her department’s operations was very comprehensive, but we’d like to zero in on the DGT’s impressive efforts with its eTranslation, which is its machine translation service dedicated to public administration.

It boasts over 1.5 billion sentences in multiple parallel languages, drawn from over two decades of work by human translators. It also has some of the most stringent security and privacy features, with translations done behind the Commission’s firewalls.

Is MT good enough for specialized professional use?

Two of the panels we attended discussed the viability of machine translation as a tool for different professional domains. There was a third which we were unable to attend, but which we include as a special mention.

The first was Edyta Źrałka’s “Human Metrics in the quality assessment and post-editing of Google Translate output in legal translation” which, as the title implies, focused on MT’s performance and usability in translating legal texts.

The other was Panagiota Frytzala’s “A Case Study on the Effectiveness of NMT for Marketing Texts”, which involved the translation of a client’s website. Frystala mentions having also translated legal text for the website, a task at which NMT excelled. However, when it came to marketing text, NMT still fell short of expectations due to the need for creativity in language.

Special mention goes to Marianna Carlucci and Samuel Läubli’s “Is Generic MT Useful for Translating Domain-specific Texts? A Productivity Study in the Ophthalmology Domain”, which also takes up the same theme.

Human vs machine, with a twist

Another interesting panel on this day was “Approaching NMT from the perspective of human translation techniques. What are the differences?” by Pilar Sánchez-Gijón and Leire Gar-Bailo. Instead of comparing the quality of machine translation vs human translation, the researchers compared them in terms of the techniques that they utilized. Naturally, human translation involved a wider variety and use of different translation techniques, while literal translation predominated for MT.

Is machine translation in conflict with copyright?

The second keynote speech of the day tackled the hard question of copyright as it relates to machine translation: “Usage rights of language data in machine translation”, by Mikel Forcada.

The mining of text for the use of creating corpora is generally protected under the principle of fair use. In addition, by nature of how MT systems treat language, it’s impossible to reconstruct any original text in any way that can come into conflict with copyright.

The hard question is what this means for translators, if their work can be repurposed to create unforeseen value through machine translation. That is, how can they benefit from this and be justly compensated for the work? It’s a question that can only be answered through unionizing and collective bargaining on the part of translation professionals.

Gender issues in machine translation

Gender and inclusivity is a topic that is gaining a lot of interest in the MT field as the researchers work to address linguistic biases that crop up in the development of MT systems. Four panels were dedicated to this very topic; we were unfortunately unable to cover them, but we’ll list them down here as special mentions:

“Neutralising for equality All-Inclusive Games Machine Translation: The All in GMT Project” by Sarah Theroine et al.

“Inclusive Language In Translation Technology: Theory and Practice, the Case of Greek” by Maria Tsigou and Valentini Kalfadopoulou

“Gender-fair (machine) translation” by Manuel Lardelli and Dagmar Gromann

More here: Dealing with gender bias in MT: An interview with Manuel Lardelli & Dr. Dagmar Gromann

“Gender Bias in Machine Translation: A Case Study in English, French, and Greek” by Eleni Tziafa