Machine translation software has gained popularity in recent years in response to the high demand for linguists. Today’s small businesses report their number one global growth goal is expanding into new international markets, so adapting their content to those markets is a top priority. Machine translation (MT) is an overall ideal solution for meeting this goal, but if used as a standalone service, will pose many risks.

MT does offer many benefits though, and as a result, the market’s compound annual growth rate is estimated to exceed 7% through 2024. However, there are losses in quality and the overall customer experience to consider. Professional grade translation may be more expensive, but it will pay long-term dividends.

Common Issues With Machine Translation Software

There are two major limits to MT regarding code and context that can easily create problems for your project. Inappropriately managing the code during an MT project can damage the user experience. And since machines cannot understand context, translations may come out differently than intended.

Parsing, segmentation, and regular expressions become problematic when making the transition via MT, particularly when it comes to structured files. Every file comes with a specific type of encoding and accompanying characters. When you run that file through MT, there’s a chance the engine won’t understand that coding.

For example, take the apostrophe. In English, it marks contractions and possessive words. However, when you change the language from English to French, it takes on a completely different responsibility. In French, it’s used to prevent the pronunciation of two consecutive vowels. When the content is run through MT, you’ll discover serious coding issues because of this discrepancy. Formatting is another major problem to consider. If you use a complexly formatted program like InDesign, the MT likely won’t understand the formatting. As a result, the content will not appear as intended and will require reworking (if not a complete overhaul).

When it comes to context, the work becomes even more complicated. There are many words in the English language that change by their industrial use. Consider the word “equity.” A simile for the term could be “fairness.” However, if one is talking about the financial industry, they’ll most likely be referring to “equity” in terms of the value of shares. An MT would not pick up on the nuance between the two.

To make MT work, you need to take some additional steps. This prep work will allow you to use machine translation software to its full advantage without being subjected to its many pitfalls.

Preprocessing and Postprocessing to Improve MT Accuracy

When using MT, you need to take some additional measures to ensure accuracy and preserve the code. These steps break down into two distinct categories: preprocessing and postprocessing.

Preprocessing

Preprocessing involves managing your code and training your MT to ensure its accuracy. Preparing the MT involves providing it with your own branded content and corporate lexicons to give it the data needed to make better decisions. Segmenting, parsing, and helping the program understand your regular expressions are required parts of code management before the MT. With better preparation you can improve the initial return, but you’ll still have a lot more work to do after it’s complete.

Postprocessing

While preprocessing is limited to managing your code and translation technology, there will be a lot more work involved in postprocessing. At this stage, you need to resemble your file and prepare it for its final destination. The collaboration of many departments is required to successfully go through four key steps:

  1. Professional linguistic checks: The eyes of actual fluent speakers will help you discover context errors that your MT will likely miss. Through this step, you get your content closer to a professional-grade translation while ensuring a consistent brand and voice.
  2. Layout: Layout can be tricky when considering how the translation will expand or retract based on your content. For example, Thai has an 80% language dilation when compared to English. To combat this, the text boxes will need to be manually fixed to ensure there isn’t much white space or strange page breaks.
  3. Workflow: Depending on the type of content, the workflow may need to be reviewed by someone in your legal, marketing, or technical departments for accuracy. In this step, you’ll need a workflow in place to take the content through these individuals before publication.
  4. Quality assurance (QA): This step involves having someone review the entire file to check its accuracy and user experience. They’ll need to compare it to the original to ensure consistency.

Machine translation software is an excellent tool, but it’s not an all-encompassing solution. There are many steps along the way that could create time sinks, thereby minimizing your return on investment (ROI). To make the most of your MT, utilize a localization management platform to help guide you through all the steps. Through this strategy, you gain efficiency and convenience while minimizing the actions that lead to wasted time and lost revenue.

Bureau Works uses machine translation software as a small component of our total comprehensive localization management. To discover our solutions, contact our team.

Published On: December 17th, 2020 / Categories: Localization Strategy, Platform Technology, Tips & Trends /

Book a Free Personalized Demo

Find out what cutting-edge companies are doing to improve their localization ROI

Add notice about your Privacy Policy here.