Machine translationTraduction automatique
How machine translation works, what today's systems still get wrong, and how systems are compared.Comment fonctionne la traduction automatique, ce que les systèmes actuels ratent encore, et comment les systèmes sont comparés.
- What to expectÀ quoi s'attendre
This module explains what a machine translation system does and why translation cannot be done by swapping words, how the two earlier generations of systems worked, and how today's neural systems and large language models translate. It then shows how systems are compared in the annual WMT shared task, and what the latest editions and benchmarks found about current systems. What machine translation and AI change for the translation professions is treated in the module on critical issues.This module explains what a machine translation system does and why translation cannot be done by swapping words, how the two earlier generations of systems worked, and how today's neural systems and large language models translate. It then shows how systems are compared in the annual WMT shared task, and what the latest editions and benchmarks found about current systems. What machine translation and AI change for the translation professions is treated in the module on critical issues.
- PrerequisitesPréalables
The page on the four paradigms, in the foundations module, gives the frame the first chapter builds on, and the chapter on attention and transformers explains the mechanism behind neural translation. Neither is required to follow the pages. No mathematics is needed. How the quality of a translation is measured is a chapter of the evaluation module, which this module points to.The page on the four paradigms, in the foundations module, gives the frame the first chapter builds on, and the chapter on attention and transformers explains the mechanism behind neural translation. Neither is required to follow the pages. No mathematics is needed. How the quality of a translation is measured is a chapter of the evaluation module, which this module points to.
- StructureStructure
The first chapter is the introduction: the task, the four generations of systems, and the places where a translator uses machine translation. The second covers the rule-based and statistical generations, and can be skipped. The third is about today's systems: neural translation, large language models, how to prompt one for translation, and what still breaks. The fourth describes the WMT shared task and what it found.The first chapter is the introduction: the task, the four generations of systems, and the places where a translator uses machine translation. The second covers the rule-based and statistical generations, and can be skipped. The third is about today's systems: neural translation, large language models, how to prompt one for translation, and what still breaks. The fourth describes the WMT shared task and what it found.
What machine translation isQu'est-ce que la traduction automatique
- What a translation system has to doCe qu'un système de traduction doit faire TRA6008 To read for S7À lire pour la S7
- Four generationsQuatre générations TRA6008 To read for S7À lire pour la S7
- Machine translation in a translator's workLa traduction automatique dans le travail du traducteur TRA6008 To read for S7À lire pour la S7
- Builds onPréalables
- The AI landscapeLe paysage de l'IA
- RelatedEn lien
- Translation memoriesLes mémoires de traduction · Evaluating machine translationÉvaluation de la traduction automatique · Neural translation and large language modelsLa traduction neuronale et les grands modèles de langue
2 concepts introducedconcepts introduits
machine translationtraduction automatique post-editingpostédition
Before neural networks: rules and statisticsAvant les réseaux de neurones : règles et statistiques
- Rule-based translationLa traduction à base de règles TRA6008 To read for S7À lire pour la S7
- Statistical translationLa traduction statistique TRA6008 To read for S7À lire pour la S7
- Builds onPréalables
- What machine translation isQu'est-ce que la traduction automatique · From rules to machine learningDes règles à l'apprentissage automatique
- RelatedEn lien
- Neural translation and large language modelsLa traduction neuronale et les grands modèles de langue · Wordfast: alignmentWordfast : alignement
Neural translation and large language modelsLa traduction neuronale et les grands modèles de langue
- Neural translationLa traduction neuronale TRA6008 To read for S7À lire pour la S7
- Strengths and limits of neural translationForces et limites de la traduction neuronale TRA6008 To read for S7À lire pour la S7
- Large language models as translatorsLes grands modèles de langue comme traducteurs TRA6008 To read for S7À lire pour la S7
- Prompting an LLM for translationRédiger une requête de traduction pour un GML TRA6008 To read for S7À lire pour la S7
- Remaining failuresLes échecs qui subsistent TRA6008 To read for S7À lire pour la S7
- Builds onPréalables
- What machine translation isQu'est-ce que la traduction automatique · Attention & transformersAttention et transformeurs
- RelatedEn lien
- The WMT shared taskLa tâche partagée WMT · Evaluating machine translationÉvaluation de la traduction automatique · PromptingFormulation de requêtes · BiasBiais
3 concepts introducedconcepts introduits
encoder-decoder modelmodèle encodeur-décodeur sequence-to-sequence modelmodèle séquence à séquence beam searchrecherche en faisceau
The WMT shared taskLa tâche partagée WMT
- The WMT general translation taskLa tâche générale de traduction de WMT TRA6008 To read for S8À lire pour la S8
- The trend across editionsLa tendance d'une édition à l'autre TRA6008 To read for S8À lire pour la S8
- Builds onPréalables
- What machine translation isQu'est-ce que la traduction automatique
- RelatedEn lien
- Evaluating machine translationÉvaluation de la traduction automatique · LLM leaderboardsLes classements de GML · Evaluation foundationsFondements de l'évaluation