1

2

3

4

Machine translation and how to make it work for your company


The growing demand for information and data to be available globally means that businesses are having to create, manage, translate and localise large volumes of documents and content promptly.

These needs are being increasingly met by corresponding technological advances in automated editing software, foreign language management and content management systems which are specifically designed to reduce the time required to process large-scale projects.

Within foreign language management, machine translation is just such a technology. It is becoming increasingly more refined and makes it possible to meet the ever-growing demand for localisation more efficiently.
 

The machine translation market continues to change rapidly as new providers and solutions constantly emerge that often deliver exponentially better results than those frequently launched just a few months earlier. This is where we come into play in making sure that your language needs are met using the most suitable product on the market today. This means that we fully integrate machine translation into our digital workflow. As is common practice, we also complement machine translations with existing translation memories and have our machine translations post-edited by subject and language experts. On request and if required, we are also able to offer the same quality controls and standards for machine translation as for our human translations. Please contact us for more information on whether MT might be a solution for your language needs.

When is Machine Translation the right solution?
  • Shorter turnaround times
    When you need information to be translated quickly.
  • Budget considerations
    If you want to decrease rather than increase the costs for your translations despite
    an increase in the volume of documents to be translated.
  • Making content accessible for information, effective decision-making
    or for large volumes of text

    When you need a draft translation or the ‘gist’ of large volumes of text to aid decision-making.

MACHINE TRANSLATION ACCURACY BY LANGUAGE PAIRS:

These scores are based on a comparison of 23 machine translation engines.
Source: https://www.slideshare.net/KonstantinSavenkov/state-of-the-machine-translation-by-intento-stock-engines-jan-2019


THE PROCESS

1. Information and advice

The options available for integrating machine translation into your language management and translation process depend on the type of documents your company uses.

2. Document type and requirements analysis

  • General information (making information accessible quickly – light post-editing or gisting).
  • Instructions (product descriptions – understanding is key – full post-editing)
  • Marketing texts (quality more important than speed – copywriting). The more creative the content, the more
    risky a machine translation.



3. Defining the purpose of specific contents

In order to determine whether your content is suitable for machine translation, it is important to establish its exact commercial purpose.

This then allows us to identify the right machine translation engine based on your quality and accuracy requirements and, if necessary, train it for specific document types and language pairs. This will require corresponding volumes of text and the provision of clean data. If in doubt, please don’t hesitate to contact us for more detailed information.

4. Selecting the best machine translation engine for you

5. Pilot testing and analysing the results

  • We would be happy to use your texts to compare and optimise different translation
    engines for you. This will involve evaluating the engine(s) currently most suited to
    your needs based on the required translation quality, speed and languages.

  • Apart from creating translation memories (e.g. from the post-edited versions of your
    machine-translated texts) we are also able to customise the machine to your
    terminology and text style. Some translation engine models can also be enhanced or
    customised through training. Please don’t hesitate to contact us for more customised through training.
    Please don’t hesitate to contact us for more information.

6. Selecting the best machine translation engine for you

Some translation engines are more suitable for some language pairs or document types than others. Thanks to our links to a number of different machine translation service providers, we are able to work out the best service available in the market for you and your translation needs.

7. Creating a post-editing programme (PEMT)

Post-editing is the process by which machine-generated translations are amended by a translator/post-editor to give them the quality of a human translation. Depending on a translation’s intended use, translations can be post-edited in three different ways: :

- Full post-editing

The result is a fluent, easy-to-read target text in which the grammar, syntax, punctuation, style and terminology have been corrected. The translation will accurately reflect the source text. This is an option where the time or budget available do not allow for a higher-quality human translation. The post-editing is performed in line with a style guide and to specific quality criteria.

Full post-editing involves e.g.:

  • Not over- or under-editing
  • Avoiding preferential changes
  • Not changing the MT style and not creating a new translation
  • Correcting all machine translation errors
  • Checking for spelling, grammar and mistranslations

- Light post-editing

Following machine translation, the text will be checked by a native speaker to make sure that it is correct and makes sense. This does not involve checking the grammar and style. This type of post-editing is most suitable for texts intended for internal use or where large volumes of information need to be translated in a short period of time.

- Gisting

This option involves no review or correction by a human of the output generated through machine translation.
This is because the information is only needed to gain a rough understanding - the ‘gist’ - of the content of the document.

8. Optimizing source texts for machine translation

The output produced by translation engines greatly depends on the quality of the source text.

That is why technical writers should always write company texts with a view to global publication. That is, in order for machine translations to be successful, the associated source texts need to be standardised and adapted for machine translation, be consistent in their use of terminology and use controlled language.

This is why using authoring tools during the writing process is an effective way to prepare for the MT process.