Knowra Machine translation Machine translation Machine translation uses computer systems to translate text or speech from one human language into another. Its methods range from hand-written linguistic rules to models trained on large collections of translated text.
Neural machine translation : Machine translation that uses neural networks to model and generate translations. Its neural models are the dominant approach in many modern translation systems.
Parallel corpus : A collection of texts and their translations aligned across two or more languages. Parallel examples teach data-driven systems how source and target expressions correspond.
Georgetown–IBM experiment : A 1954 demonstration that used a computer to translate Russian sentences into English. It brought early machine translation research to public attention.
Computer-assisted translation : Human translation supported by software such as translation memories, terminology tools, and text editors. It keeps the human translator in control, unlike fully automated machine translation.
Translationese : Linguistic patterns in translated text that differ systematically from originally written language. Machine-generated translations can reproduce or amplify these distinctive patterns.
Statistical machine translation : Machine translation that estimates translations from statistical patterns in bilingual data. It replaced many rule-based systems before neural methods became widespread.
Tokenization : The process of dividing text into units such as words, subwords, or characters. Translation models operate on tokens, and token boundaries affect vocabulary and output quality.
Warren Weaver : An American mathematician and science administrator who helped frame early ideas about mechanical translation. His 1949 memorandum helped launch sustained research into machine translation.
Speech translation : The automatic translation of spoken language, often combining speech recognition, translation, and speech synthesis. It extends machine translation from written text to spoken conversations.
Hallucination (artificial intelligence) : An output generated by an artificial intelligence system that is unsupported by its input or reliable evidence. A fluent translation can add, omit, or distort information present in the source.
Show all 27