hard-drive Exploring automated translation

Automated translation tools can often do a "good enough" job of translating between languages, with some differences.

Many technical writers need to create content for an international audience. For years, the industry best practice has been to hire a bilingual expert to perform not just translation but localization of the content for the language and culture of the intended audience. Translating content means converting the material from the source language to the target language, such as translating a text written in English to Spanish. Localization or internationalization takes it a step further, and includes adapting the text to suit the cultural differences of the audience.

Technical communicators now have access to a variety of automated tools that can provide translation of text from one language to another. Google Translate and Microsoft Bing Translator are popular tools with good reliability; for many texts, Google and Bing do a "good enough" job of translating works between languages.

I wanted to explore how well these tools can translate material. For this experiment, I wrote a sentence that might come from instructions for a made-up device called the VCX 100. I wrote three versions of the sentence: one that was constructed at a "high" or "very formal" reading level, another at a "moderate" level, and a third at a "low" or "overly-familiar" reading level. For each version of the sentence, I used Google Translate and Bing Translator to translate the text from English to Spanish, and back again. I chose Spanish because I have some fluency with Spanish, and I feel I can better evaluate the results.

High level

Original text: In this endeavor, we iterate several possible configurations of the VCX system model 100 and appropriate recommendations for correct operation of the same.

Google Translate

This sentence uses very formal language to introduce the VCX 100, using very long words with a more complex sentence structure. Using Google Translate gives this text: En este esfuerzo, iteramos varias configuraciones posibles del sistema VCX modelo 100 y recomendaciones adecuadas para el correcto funcionamiento del mismo.

Using Google Translate to "reverse translate" back to English gives basically the same text: In this effort, we iterated several possible configurations of the VCX model 100 system and appropriate recommendations for the correct operation of the system.

Bing provides essentially the same text when translated back to English: In this effort, we iterated several possible configurations of the VCX Model 100 system and appropriate recommendations for its proper operation.

Bing Translator

Bing translates the text into Spanish in this way: En este empeño, iteramos varias configuraciones posibles del sistema VCX modelo 100 y recomendaciones adecuadas para el correcto funcionamiento del mismo.

Using Bing to "reverse translate" back to English provides basically the original text: In this endeavor, we iterated several possible configurations of the VCX model 100 system and appropriate recommendations for its proper operation.

Using Google to translate this back to English provides substantially the same text: In this endeavor, we iterated several possible configurations of the VCX model 100 system and appropriate recommendations for its correct operation.

Comments on translation

The main difference between the Google Translate version and the Bing Translate version is how "endeavor" is translated. In Spanish, the noun esfuerzo means (in physics) "force" such as "how much force is needed to move this?" or (in engineering) "stress" as in "shearing can cause stress on a building." In both cases, esfuerzo really means "something that happens" but in a physical application. However, the noun empeño can mean an undertaking, such as something that you do. A native speaker would likely translate "endeavor" as empeño.

Medium level

Original text: Follow these instructions to set up and use the VCX 100 device.

Google Translate

This is a short phrase written at a more "moderate" reading level. It is more direct and uses action-oriented language. Using Google to convert this to Spanish gives this translation: Siga estas instrucciones para configurar y utilizar el dispositivo VCX 100.

Google translates this back into English using the original phrasing, with the addition of the word "device" at the end: Follow these instructions to set up and use the VCX 100 device.

Using Bing to "reverse translate" the Spanish translation to English gives the same text as the original: Follow these instructions to set up and use the VCX 100.

Bing Translator

Interestingly, using Bing to translate the text into Spanish results in the same text as Google's translation: Siga estas instrucciones para configurar y utilizar el dispositivo VCX 100.

And not surprisingly, using Bing to "reverse translate" into English gives the same output as Google: Follow these instructions to set up and use the VCX 100.

Using Google to translate back into English is the same: Follow these instructions to set up and use the VCX 100 device.

Comments on translation

At the Medium level, both Spanish translations are the same, as are the translations from Spanish back to English. This seems to support the general recommendation from other technical communicators to adopt a more "medium" level of communication, and direct action wording. Doing so provides effective communication not just in native English text, but also delivers more reliable cross-translation using automated tools.

Low level

Original text: Here's how to run the VCX100.

Google Translate

This phrase is both very terse and too informal. This style of wording might be more appropriate for an email or text, but not in written documentation. Using Google to convert this short text into Spanish gives an equally brief translation: Así es como se ejecuta el VCX100.

Perhaps because the text is too brief, Google has some difficulty in "back translating" the text from Spanish into English. The generated text is somewhat awkward: This is how the VCX100 runs.

Similarly, Bing translates the terse phrase back to English with awkward phrasing: Here's how the VCX100 runs.

Bing Translator

Bing also has some difficulty in effectively translating the phrase into Spanish: A continuación, se explica cómo ejecutar el VCX100.

However, Bing "reverse translates" the translation back into English more effectively than Google: Here's how to run the VCX100.

Google translates the Spanish back to English in the same way: Here's how to run the VCX100.

Comments on translation

This brief statement results in several differences in translation. For example, the word así means "So," such as a continuation in a sentence like "And so, we are there." The word continuación means literally continuation, so "A continuación" means "to continue." Both are interesting ways to start a translation of the very informal "Here's how…"

Bing Translator instead uses the verb explicar, which means "to explain." The phrase "se explica" basically means an explanation to you, the reader.

Both of these translations seem technically correct but it's interesting to note the different word choice in both.