Thread by @enFRdansletexte

Colloque Tralogy III #tralogy3
« #Traduction humaine et traitement automatique des langues : vers un nouveau consensus ? »
‘Human #translation and natural language processing: Forging a new consensus?’
2e journée
#TAL

Technologies de la #traduction et dynamiques de pouvoir
Confining Machine Translation; Reclaiming Localization
by Alan Melby, of Brigham Young University @BYU and the International Federation of Translators @fit_ift
#machineTranslation #localization

Translation grades
- high grade: #professional #translation
- low grade: supports triage

Translation grades are tied to risk: high risk text requires high grade translation
Confining unedited #MT to low grade

"Let's be realistic, in today's world most translations produced are low grade and they're used and useful."
"Currently unedited MT is low grade. It doesn't guarantee serious mistakes but serious-error risk is guaranteed: translation mistakes appear randomly."

"Proposed approach: #consumer right-style label showing the method of production.
We propose that once a method of production has been selected, whoever claims responsibility for the #translation (translator, agency, etc) applies a label, an icon to the translation."
#labels

"#FIT position paper on #localization:
Often it's implied that localization and translation are different or don't overlap. Here's a graph that show their close relationship. The translator is at the heart of localization."

La place de l’intervention humaine dans la #formation des traducteurs
Le projet européen MultiTraiNMT : Un premier retour sur les usages et les besoins au sein des formations universitaires françaises
@MultiNmt #traduction

Le projet MultiTraiNMT a 3 livrables :
- un livre 'Machine Translation for Everyone'
- des activités en ligne H5P
- une plateforme (MutNMT)

La plateforme MutNMT, ouverte, disponible sur Github

« La plateforme MUT peut servir d'outil d'initiation à la traduction automatique neuronale. » #TAN

Liens vers les livrables du projet MultitraiNMT

La place de l’intervention humaine dans la formation des traducteurs
Is it difficult to launch a pioneering Master's program in Translation?
Maya Lyutyanskaya, of Northern Arctic Federal University @NArFU_GlobaL

"Master's degree program on computer-assisted and machine translation
The master's program is the first of its kind in Russia."

"Collaboration with other further education institutions
- in social sciences, humanities and #international #communication
- in IT and automated systems
- in natural sciences and technologies"

"Skills and competences students learn:
- linguistic (Russian and 2 foreign languages)
- translation theory
- editing and QA
- IT: HTML and mobile and web development
- psycholinguistics
- business coaching"

Q for Alan Melby: label idea: how to deal with the many shades of part-MT? Add a new metric?
A. Melby: "After debate, we decided we want a very binary labelling system: unedited MT and all the other shades of human intervention."

"The best thing for the consumer is not to get lost in this spectrum. Either there's just a machine or there's a human involved, taking responsibility for the translation."

Q for Maya Lyutyanskaya: do you already see changes in the professional integration of your students?
M. Lyutyanskaya: "The company that took part in creating the professional standards was the one that raised the issue of lack of properly trained professionals working with MT."

"It's a recent program so it's probably too early to assess the results but we are following the needs the industry tells us it requires."

Alan Melby "We are open to suggestions about the letters present on the label. Σ suggests, subtly, that translation is about an algorithm, whereas Ω hints at human intervention."

Q for Maya Lyutyanskaya: how did you choose the two foreign languages (English and German) your students learn?
M. Lyutyanskaya: "English, German and French are the languages traditionally taught at our university. There was also the aspect of requests by our students."

Q about the MutNMT platform: where does the computation happen?
"The server in Barcelona has been bought for the project but regarding the Grenoble instance, I think it's on the cloud."