Building and research
Building and research about machine translation
Machine translation is the task of translating natural language automatically. The input is content in one language, and the output should be the same content in another language.
The most common input and output type is text. There are also other input and output types, like speech.
Most machine translation systems translate each sentence separately, without any context. Most machine translation models support translation for only one language and in only one direction.
The most common approaches use human translations as training data.
Table of contents
- Approaches
- Data
- Training
- Quality evaluation
- Metrics
- Multi-engine machine translation
- Tags and placeholders
- Locale
- Quality estimation
- Other input types
- Bridging
- Human-in-the-loop
- Zero-shot translation