Machine Translation vs Human Translation

Machine translation is growing both in terms of its quality as well as use by customers. Although the quality of machine translation improves with advancements in artificial intelligence, it is clear that it does not match the complexity of human thinking and lacks the ability to discern idiomatic, cultural and emotional nuances.

What is Machine Translation?

Machine translation (MT) is translation by a machine. Translation applications such as Google Translate and Babelfish use artificial intelligence (AI) to convert in a given language to other languages.

AI uses combinations of dictionaries (embedded memory), algorithms and statistical methods. Rules of syntax that are coded into the system help identify the source and generate the target sentences.

What is Computer Aided Translation (CAT) and Translation Memory (TM)?

A Computer Aided Translation Tool, on the other hand, is software that is used by a human translator in automating their work. This is not a machine translation but only a tool that assists the translator in creating a translation memory and term bases, among other things. It facilitates ease of translation by dividing text into logical segments, creates a view of a segment and a space to type, thus facilitating the translation process.

What is the difference between MT and CAT?

A CAT tool does not translate by itself, but may throw up suggested translations based on the memory created by the preceding translations. It is just like we use a word processor for writing and a spreadsheet for crunching numbers. They are tools that ASSIST humans in translation. MT on the other hand is an artificial intelligence tool that suggests translations from its repository, without human intervention.

The line between CAT and MT is getting more and more blurred as MTs are now beginning to add the translation memories from CAT tools to their repositories. So CAT tools are becoming the basis for data collection that feed MT. Over the last decade we have seen a tremendous improvement in the quality of MT and the trajectory continues in the positive direction.

Advantages of Machine Translation

Machine translation is correct (with reservations) to some extent, as far as factual statements are concerned. It is good for literal translations.

MT is consistent, fast and cost effective. However, the limitations seem to outweigh the advantages.

Limitations of Machine Translation

MT is unable to translate idiomatic language. More often than not, it throws up literal translations of figurative speech. This can lead to major embarrassment.

AI does not understand emotion, tone and context. We all know that language is more than the sum of the meaning of words. AI can only recognise words but not the overall sense of a sentence, nor the intention, mood and context. As we say, it often misses the point.

MT does not understand cultural differences. It only recognizes linguistic elements and converts them to the target language, rather than translate.

From a linguist’s perspective, a machine translation usually proves itself at the level of syntax and lexical correctness, but fails at the semantic and pragmatic level. This means, language can recreate grammatical structures in either language, but will often fail to make the right judgment where a word has several meanings. It fails to see the context, function and purpose of the said word or phrase. (Ignore this paragraph if you are not a linguist or do not understand the jargon

Examples of Errors in Machine Translation

Below are some example of laughable errors that machines make while ‘translating’.

When is MT OK to use?

Below are some example of laughable errors that machines make while ‘translating’.

When is it NOT OK to use MT?

It is NOT OK to use MT for publication or consumption of external parties without a round of human check. Post editing is absolutely necessary. If you blindly rely on MT, your document will be derided or your website will become an international joke! Legal documents should BEST be edited by humans, and technical translations require it too. Literary texts should NEVER be translated by machines.

Conclusion

Although the scope of machine translation is expanding and it continues to add value, the importance of human intervention and post editing can never be overemphasised. A machine has limited intelligence. Human thinking is far more complex and has many dimensions, which are difficult for machines to scale.

Choice of a Translation Agency for Natural Human Translation

Human translation is therefore the way to go. Be it in the form of original human translation or post editing, which may save cost to an extent. It is essential that you identify the right translation and localization company that will make sure your message is retained in translation.

WordPar international is a leading translation company offering services in India and the world. Office of the company are present in Bangalore, and virtual offices in Delhi, Mumbai and Chennai.