Imagine describing a funny movie to your friend. Now, imagine describing that same movie to the president of a large corporation. Would you use the same language? It’s unlikely. Humans are skilled at gauging when to use formal language vs. informal language. Machine Translation (MT) engines are generally unable to do so — unless they have help.
It’s imperative to be aware of this MT limitation and take steps to overcome it to develop a strong relationship with your target audience.
Language register or style defines the way we use a language according to sociocultural context, customs, and communication channels. In other words, it is the degree of formality with which we express ourselves. As noted, we do not speak the same way to a friend as we do to the president of a company.
Levels of formality — formal vs. informal language — are fundamental in defining the tone of a given discourse. Different writing styles, idiomatic expressions, pronoun variations, and verb morphology may have different levels of formality depending on the language, country, social group, or professional environment.
When writing content, ask yourself the following questions to establish the appropriate level of formality:
How do I want them to feel when using my product?
Do they have a specific profile?
When translating content into different languages, customers often create style instructions that include a profile of the target audience, the context of the project, and the expected tone and level of formality.
It is difficult for MT engines to produce consistent formality in the target language. It is especially challenging when the source language has fewer formality levels than the target language.
MT models typically return a single translation for each input segment. When the input segment is ambiguous, the model must choose a translation among several valid options, regardless of the target audience.
Letting the model choose between different valid translation options may result in inconsistent translations or translations that have an incorrect level of formality.
It’s imperative to achieve the correct formality style in your translated content.
Speakers from certain cultures may perceive an incorrect level of formality as rude. Additionally, there are many cases in which formal language may simply be wrong, not because it is offensive but because it is totally out of place. Errors like these can alienate your audience. For example, translations for computer games and student programs require a casual tone.
When companies take the time and make an effort to adapt the appropriate style to the target during translations, they are sending a strong message that they care about their customers. Such efforts may be even more appealing to potential prospects. Effectively connecting with your audience helps you to succeed.
It can be difficult to match the formality level of the target language to the audience during translation and especially when using MT. Problems can arise when translating content from English into languages that have multiple levels of formality expressed through a grammatical register.
How formality distinctions are expressed grammatically and lexically can vary widely by language.
There are languages that have well-defined formal modes, such as Spanish, which separates "tú" from "usted," German "du" from "Sie," French "tu" from "vous," and so on. Even in Chinese, whether traditional or simplified, this differentiation exists. Korean has at least six levels of formality. In Japanese and other languages, the distinctions expressing polite speech may be morphological markings on the main verb, as well as on some nouns and adjectives, specific lexical choices, etc.
In other cases, as in English, there are no pronouns to mark these distinctions, so we must rely on the context to express this idea.
This creates an additional problem for most MT engines, which are inconsistent when differentiating between formal/informal language styles.
Several strategies enable MT systems to generate a range of formality styles.
Rules-based techniques can produce consistent Machine Translations in the desired style or formality. These rules reliably replace elements that determine the undesired style with a correct translation returned by the third-party system while preserving the meaning.
Non-rules-based techniques may be used to develop custom Machine Translation models that consistently produce translations in the desired style or formality. This approach uses supervised training and annotating politeness in the corpus set to obtain the politeness feature.
The goal is to obtain a single system trained with diverse data that can adaptively produce results for a given formality style.
Most MT systems do not support language formality or gender parameters. However, some engines offer this support, and we expect greater demand for these features.
At present, DeepL (API) and Amazon (console and SDK) offer features that control formality during translation for a limited number of languages.
There are three options to control the level of formality in the output: default, formal and informal.
The default option does not change the formality of the Neural Machine Translation (NMT) output. The formal/informal function allows the user to choose between a formal or informal tone of voice. Specifically, the function sets the pronouns and related words used in the translation.
Lionbridge’s Smart MT™, an enterprise-grade Machine Translation solution, allows linguistic rules to be applied to the target text to produce Machine Translations with the desired style or formality.
Our specialists maintain an updated database of rules that are fed back into the analysis of MT outputs to control the outcome.
Provided that there is sufficient material, we combine a rule-based strategy with custom Machine Translation models to achieve an optimal result.
If you’d like to learn more about Lionbridge’s next-generation Machine Translation technology and how Lionbridge can help you fully capitalize on Machine Translation, contact us today.
