Post-editing machine and AI-generated translations
We help companies that use machine translation recover as much of the quality lost through MT as possible
Machine translation (MT) is increasingly in demand among businesses because it is a low-cost option that delivers rapid, approximate translations. However, the quality of the resulting output is often inadequate.
SMG Languages works with native-speaker translators who specialise in post-editing MT and AI-generated translations.
In practice, machine-translated output can be regarded as a ‘semi-finished product’: a form of machine-generated language with conventions and characteristics that may resemble natural language while remaining fundamentally different from it.
Machine translation must therefore be post-edited to produce a finished text that can be used without the risk of serious inaccuracies and errors commonly associated with MT.
SMG in the era of new technologies and AI
In line with our corporate Vision, the primary objective of language services is to break down the barriers that prevent people from communicating. These services enable and facilitate dialogue between individuals separated by language, culture and/or traditions.
Language services must preserve the focus on human relationships: between the sender and the recipient of the content, as well as with the mediator. Mediators apply their linguistic and socio-cultural skills, together with their knowledge of both parties, to ensure that the message is conveyed clearly and completely, including its explicit and implicit meanings.
Limits of machine translation and AI
New technologies such as machine translation and artificial intelligence (AI) can help improve and facilitate communication between individuals. However, as tools rather than people, they cannot reliably grasp every layer of meaning in a dialogue. Communication consists not only of written or spoken text, but also of non-verbal communication and implied concepts. These context-dependent elements cannot always be fully interpreted by a machine.
In the field of linguistics, these powerful tools can support the work of language professionals. However, the human role of certifier, guarantor and intermediary must be preserved—a role that professionals have always fulfilled and that machines cannot replicate—to safeguard the meaning and purpose of multilingual communication.
Why does machine translation require post-editing?
First, machine translation attempts to reproduce the human communication process, but it cannot fully analyse and understand the depth of human thought. Consequently, some of the information intended by the writer may be lost in the target text.
For example, MT software cannot independently distinguish between terms that are considered synonymous, even though they may carry subtle differences in meaning and evoke different responses in readers, including as a result of cultural and cognitive biases.
Another example is figurative language: machine translation software is often unable to recognise, understand and adequately translate idioms, proverbs, aphorisms and popular sayings. These may therefore be translated literally, without taking into account their extralinguistic meaning.
Other examples include metaphors, expressions involving irony and sarcasm, dialect and slang terms, and neologisms.
For these forms of expression commonly used in human communication, fully automatic translation has not yet achieved consistently satisfactory quality. For this reason, translations produced using automated tools almost always require revision, especially documents and texts intended for use in a business environment.
Machine translation post-editing: why it is an innovative service
Revision and post-editing date back to the earliest days of the translation profession. They have always been considered essential, particularly in fields such as politics, economics and law, for checking that a translation is faithful to the original.
Nowadays, we believe that revising a machine translation is not comparable to revising a translation produced by a human translator. Editing MT means working not with human reasoning, but with machine-generated output, which may be distorted or lack genuine understanding—even when generated using AI. Above all, it lacks a human sense of ‘purpose’: the translation is not necessarily adapted to its readers and the context in which it will be used, because a machine cannot fully interpret that context.
For this reason, the editor needs to specialise in reworking texts pre-translated using MT. This means not only correcting grammatical errors and imperfections caused by misunderstandings, but also reworking the text as necessary to restore contextual information so that communication is effective and fulfils the writer’s purpose.
Get a
Free quote
Fill in the form below to get a quote from our offices. We are happy to respond to your request and will get back to you as quickly as possible.