Computer-Assisted Translation Systems Versus Machine Translation
We’re all familiar with Google Translate and have probably used it at one point or another. But do we really know what’s behind it and what the language industry makes of automated translation tools? If you’re an individual or a company stakeholder considering using a translation service for your documents or website, in particular, you should be aware of these technologies to make an informed decision. I’ve been using translation technologies for two decades in my capacity as German-to-English translator, so let’s take a closer look!
What is a CAT tool, and what is a translation management system (TMS)?
A CAT (computer-aided translation or computer-assisted translation) tool is software that supports translators in preparing translations. It converts a text to be translated, segments it, and then makes these segments available for translation in its own editor.
A CAT tool comprises several subsystems. At its core is a Translation Memory (TM), which collects individual translation units in a database. During the translation process, the individual translation units, consisting of a segment of the source text and the equivalent translation in the target language, are saved. Storing the translation units ensures that translations are consistent and that the same terms are used within a text. This improves the quality of the output text.
The CAT tool divides the source text into segments such as sentences, bullet points or headings. The individual segments are then displayed to the translator in an editor window. The translator enters their translation in the corresponding segment next to or below the source text. This creates translation units that are stored in the TM, which is a bilingual database containing both the original source content and the corresponding translation. When the translator translates a similar text in the future, the program suggests existing, saved translations.
A CAT tool or Translation Memory, therefore, does not translate independently, unlike machine translation programs, but only serves to support a human translator and stores the created translations in a database. In fact, studies have shown that translators work up to 28% faster when using a CAT tool.
Advantages of a computer-assisted translation system
CAT tools thus enable faster and higher quality translations and also ensure the consistency of those translations. What’s more, CAT tools can prepare and segment a wide variety of file formats such as Microsoft Word, Excel and PowerPoint files, Open Office files, and even XML files. This allows translators to process certain file types even if they don’t have the original software in which they were created, e.g. InDesign.
Today, CAT tools are industry-standard programs used by both translation companies and translators across the globe.
Using a CAT tool and associated Translation Memory lays the foundation for future translations – which grows with each completed translation. If a new source-language file is opened in a CAT tool and the translation memory is used, ‘new words’, ‘100% matches’ (identical segments) or ‘fuzzy matches’ (similar segments) are displayed when the file is translated. These are then checked by the translator and, if necessary, adjusted so the target segment can be taken over. As a result, the wheel doesn’t have to be reinvented every time.
Furthermore, the advantage of Translation Memory systems is clearly evident when there are high repetition rates within a document. The respective segment only has to be translated once and can then be inserted automatically in all other instances in the text by the tool. This makes translators more efficient and can save costs for clients.
A CAT tool analyses each new text by comparing it to the units stored in its Translation Memory. It then evaluates how much content there is to translate, and this information is often used to calculate costs. An analysis is typically broken down like this:
New words: segments which have never been translated before and are not present in the Translation Memory
Fuzzy matches: segments which are similar to something previously translated and are found in the Translation Memory but need to be edited by the translator
100% matches: identical segments which are found in the Translation Memory
Context matches: an identical segment that appears in exactly the same context (i.e. between two 100% matches) as in the Translation Memory
Repetitions: identical segments that occur more than once in each file; the translator only has to translate this once, and the tool automatically replicates the translated segment across the whole document
Each type of segment will require a different level of effort from the translator, from a completely new translation from scratch to a review with no changes.
When a translator uses a CAT tool for the first time and their TM is still empty, all segments will come back as ‘new words’. But as soon as more text is stored, leverage will start to increase.
In addition to the Translation Memory, CAT tools also boast many other features including:
Quality assurance features
Search and replace functionality
Text alignment functionality
Today, CAT tools are available as desktop versions, and server-based or cloud-based. Server-based versions are usually used by larger language service providers as they allow multiple translators to work on a project simultaneously.
Popular CAT tools
Some of the most popular CAT tools on the market are:
SDL Trados Studio
It’s important that we don’t confuse CAT tools with machine translations programs.
What is machine translation?
A machine translation application is a program that feeds a text to a computer algorithm which automatically translates it into another language. So no human is involved in the translation process.
Nowadays, though, many CAT tools offer a machine translation add-on. Interestingly, some translators refuse to use these on ethical grounds or because they think it will lead to a poorer translation and dilute the craft.
We differentiate between three types of machine translation methods:
1) Rules-based machine translation:
This method relies on grammar and language rules developed by language experts, and on highly customisable dictionaries.
2) Statistical machine translation:
This method does not rely on linguistic rules and words; instead, it learns by analysing large volumes of existing human translations.
3) Neural machine translation:
This method teaches itself how to translate by using a large neural network; it is becoming more and more popular because it provides better results than the other two methods. Google MT and Microsoft Translator, for example, rely on neural networks.
Popular MT tools
Some popular machine translation programs are:
Many language service providers nowadays deploy some sort of machine translation program and have the output revised by human translators or editors, referred to as ‘post-editors’, in order to improve the results, which are often subpar.
Naturally, costs are lower for these automated programs than for human translation, but the trade-off is invariably the lower quality. Language is highly dynamic and complex, and while machine translation technology has improved greatly over the years, it will never be able to accurately identify the nuances of each language in their entirety and convey them in another language like a human translator can.
So, no matter how similar they sound, CAT tools and Machine Translation programs work differently and are used for different purposes. In a nutshell, CAT tools are used by professional translators and not suitable for lay users. Human output is, of course, limited and costly. This is where machine translation comes in. Machine translation programs can be used by anyone at low to no cost but are far less reliable. They are often deployed in conjunction with human reviewers by language companies.
If you require translation services, you can now weigh up your options and make an informed decision that’s right for your needs.
Photo by Fotis Fotopoulos on Unsplash