Today, machine translation is necessary as part of the localisation process as it helps to enable translation and localisation to take place quickly and efficiently so reducing the costs of translations. As more and more resources are devoted to machine translation technology the quality of translations is markedly improved.

Definition of machine translation

Machine translations (MT) are automated translations created by computer software. The user inputs the text in its source language and then selects the target language. The MT software then generates the required translation. Machine translations may be used for quickly translating large amounts of text which would be virtually impossible using the traditional translation method. It could be used for the translation of whole texts without the need for human input (raw MT), or in conjunction with human translators.

How machine translations work

In the last ten years, there have been significant developments in MT such as neural machine translations and artificial intelligence.

Neural machine translations

Neural machine translations (NMT) are a machine translation approach constructed on neural networks. The network may be divided into two parts which are an encoder which reads the imputed sentence and creates a representation that can be used for translation and a decoder which creates the translations. Words and whole sentences are in NMT represented as vectors of numbers. NMT appraises the fluency of the complete sentence.

Machine translation training data

Machine translations use training data but generic MT engines, such as Google Translate, Amazon Translate, and Microsoft Translator are used more for general purposes and have not been trained with data that matches an exact topic or domain. Data is collected continuously and used to improve the output. However, custom MT engines are more finely-tuned as they have been trained with precise data which means a more accurate MT translation but costs more.

The advantages of machine translations

Machine translation has some significant advantages compared to traditional translation which makes it an attractive proposition for use by businesses.

• It is fast. 
• It can easily handle anything from one document to thousands.
• It is cost-effective and even up to a thousand times cheaper.

Despite the ground broken by MT, human translations, or post-editing, are still considered to be the gold standard for any translations that require perfect quality.

 When machine translation is preferred

Machine translations work for different kinds of content but if it’s creative like marketing copy, machine translations might not be the solution. It may be used, to begin with, but it is better to use human translators who have the expertise to be more creative with the translations.

You also need to be certain that the machine translation output will meet the reader’s expectations. Translated content designed to showcase your product or company should be checked by a human translator. However, if the translated content is to be used internally in a company, MT may be the best solution.

If your deadlines are tight and there is no one available to undertake a translation job, MT is a good choice. Also, low-priority translations like internal documents, make perfect candidates for MT. 

Machine translations and post-editing (human input)

To help improve the quality of MT, post-editing (PE) brings together MT with human translations. Trained linguists offer their expertise so that the best quality translations can be created using both MT and human input at fast speeds.  It is possible to choose an optimal level for MTPE depending upon your translation requirements.

Fortunately, for the translation industry, human translators will always be needed to play a key role in MT translation workflows as post-editors so that the MT output maintains consistent quality.