CAT tools vs. Machine Translation: What’s the Best Method?

To get more clients outside of Australia – specifically among non-English speaking audiences – most companies need professional translation services.

There are two similar-sounding tools which you can use to aid the process of translating your content for other audiences:

One is Machine Translation (MT). The other, Computer-Assisted Translation (CAT) tools.

Both methods require the use of computers and software, but their processes are very different. This article will discuss the pros and cons of CAT tools and Machine Translation, helping you to decide which of the two methods will suit your needs better.

Simply put, a CAT tool is a program or software that aids a translator with their work. Although the method uses a computer to organise foreign words and generate a virtual glossary of terms, the entire translation process is still done by a person. You only use the software for systematising your human translations.

The three main functions of a CAT tool are to:

1. Break the text into sentence fragments and present it to the users in a convenient manner. Some programs even provide separate sections for viewing the original and translated text.
2. Save the translated segments along with the source material and present them as translation units (TU). With navigation tools found in the software, you can easily reread your previous work and locate the segments that you need to revise.
3. Save the translated bits of information in the software’s Translation Memory (TM) or database. If you need to reuse the units for your future projects, you can then simply view them in the database. Aside from saving your time, it also helps the translator to be consistent with the terms they’re using. However, you have the option to revise these terms to suit a particular task.

CAT translation tools have been around since the Cold War. During the huge superpower stand-off, there was a great need to translate vast amounts of information in a quick manner. Intelligence services spent considerable sums of money to build technology to help them out. America and Russia were the very first countries that experimented with automatic translation programs. Some of the earlier forms of CAT tool are Systran (System Translator), and the TSS (Translation Support System).

During the 1990s, introducing these tools to the market was difficult because they were expensive and the translation industry relatively immature. But with the advent of the internet, Computer-Aided Translation software has become a necessary tool, especially in the field of technical translation.

One of the major benefits of using CAT software is that it speeds up the translation process, helping you get large projects completed faster. Plus:

• Since the translated terms and original text are both stored in a database, you can maintain consistency and quality in your translations too.
• You can also share the database with multiple translators working across different divisions and projects.
• The tools are also compatible with different file formats.

But just like any kind of software, CAT tools are still subject to error. Sometimes, what is a correct translation for a segment in one context is not appropriate in another – especially if you’re working on bigger projects. That’s why translators still need to double check even exact-match segments and update the Translation Memories as necessary.

Machine translation deals with automated translation or the use of software alone to translate words from one language to another. MT engines rely on human-made translations for their training data. Once the engine has been trained with a corpus of existing translations, it can apply its new knowledge to other texts.

MT engines use one of several different approaches for translating:

• Rule-based approach: wherein the engine will translate words based on grammar rules.
• Statistical approach: wherein the program will translate a text based on the frequency of usages in the training data.
• Deep learning approach: wherein machine neural networks are used to produce translations of a fluency heretofore unseen.

Sometimes a hybrid method which include multiple approaches will be used.

The use of machine translators started in the 1950s. The Georgetown experiment, conducted in 1954, was one of the very first projects that required the use of automated translation software. Over sixty Russian sentences were translated into English through this experiment. The project became successful and led to the funding of future machine translation studies.

Sadly, the actual development of this tool ended up being slower, despite the high hopes of researchers. After a decade of fruitless research, funds for machine translators declined. Even with today’s advanced technology, scientists haven’t created a translation tool that’s fully automatic and which can produce high-quality output consistently. However, there are many free and budget-friendly programs that provide helpful translated output.

That said, for all high visibility content, the importance of getting proper human post-editing of your MT output cannot be overstated.

Let’s start with the advantages of Machine Translation:

• If you’re pressed for time, you can always rely on these tools. You don’t have to wait to schedule in the work with human translators. Machine Translation engines can translate hundreds of thousands of words in a few hours.
• The cost per word is typically very low.
• Machine Translations can protect sensitive data by preventing disclosure to human translators.

However, there are also several disadvantages of Machine Translation:

• Even though you can save time on translating a document, an editor might consume more hours in revising the output to bring it up to an acceptable level of quality.
• Machine translations are systematic and bound to formal rules. They can be effective for documents which don’t require stylish translations, but they cannot solve ambiguity or translate a phrase based on context.

Both tools can be excellent for the right project:

MT engines are handy if you need quick translations, or you just want to give your audience an overall idea of what you mean without minding too much about grammar or style. You can gradually improve performance over time by editing the output and using the revised versions to train the engine so it can provide you with better translations next time.

But if you really want quality translations that take into account the context and culture of your target country, then you should use a human translator supported by a CAT tool. In the hands of an adept linguist, this software will produce publication-standard translations that your audience can identify and connect with.

This will be true whether you’re targeting the more than half a million Mandarin speakers in Australia, the 300 000 Arabic speakers, or a foreign audience in any country on the face of the Earth.