Neural Machine Translation Explainer
Neural machine translation, or NMT, trains one model to translate text from a source language into a target language. Unlike phrase-based systems built from separate components, NMT learns an end-to-end probability distribution over possible target sentences.
An encoder represents the source sequence, and a decoder generates the translation token by token. Attention lets the decoder refer to relevant source positions. Modern systems commonly use transformers and shared subword vocabularies.
NMT powers online translation, multilingual communication, localization, document workflows, and cross-language search. Multilingual models can share patterns across several language pairs and sometimes transfer knowledge to pairs with less parallel data.
A translation can be grammatical yet omit, add, or alter meaning. Rare terms, names, dialects, cultural context, and low-resource languages remain difficult. Professional and high-stakes use requires terminology controls, document context, human review, and evaluation by qualified speakers.
Neural machine translation uses a trained neural model to convert a sequence in one language into a sequence in another.
Key terms include neural machine translation, source language, target language, encoder, decoder, attention, subword and parallel corpus.
The encoder represents the source while the decoder predicts target tokens in context.
Attention connects each generated portion with relevant evidence from the input.
Grammar and confidence do not guarantee faithful translation of names, negation, or technical terms.
Human review remains essential for legal, medical, safety, and public-facing communication.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.
