Imagine being sat at a table with your dearest friends while you are telling them a funny story. The used language is surely different than the one you would use while telling the same story to the CEO of an important company. This is reason why, based on the context, the speaker chooses a formal language or an informal one. These differences look natural for human beings, but automated translators can not properly identify them yet.
What Is “Formal Language”?
In linguistics, the language register is a language variety that depends on the social relation between the speakers. The more the register is higher, the more the language will be formal. Just like the previous example, we do not speak in the same way to our friends or to an important CEO.
Clearly, the language register is necessary to understand the tones and it must be respected when the text is being translated. Idiomatic expressions and even pronouns, like in Italian, are elements that decline themselves in a different manner according to formalities and the register. The formal language, for example, does not include obscene language or slang.
In order to help translators to adapt a text coherently, clients usually specify:
- The original intentions of the text;
- The target profiles;
- The context of the project;
- The tones and the formality level of the source text.
Why Machines Can Not Interpret Formal Language?
Machines have difficulties in interpreting language registers, especially when formalities level do not match in the source language and the target language. These lexical ambiguities bring machines to choose between different options that often do not respect the chosen target, leading to incoherent texts.
For example, in English there is no pronoun that is used according to formalities. In Italian, instead, the pronoun Lei is preferred in a formal context instead of Tu (You). Japanese is even more peculiar since there are morphological artifices on verbs and a specific vocabulary based on context. Automated translators can not interpret this kind of contexts especially for these differences in cultures.
There Are Attempts for Forms of Register Checking in Machines
There have been several attempts to develop technologies that could help machines in interpreting different levels of formalities. This mostly happened by using machine learning strategies, with which automated translators learn to recognise various contexts. Another attempt was to establish rules that indicates the desidered formality level earlier.
However, this highlights the importance of translators in video games industry. Achieving an appropriate and formal language is a distant goal for automated translators. A non-mathematical reasoning, which leds to a rational translation choice, is still a quite human quality. Only the future can tell us if machines will be able to replace us in this kind of activities.
Meanwhile, machines can still help us with instruments like CAT Tools. Those are softwares that help translators to orient themselves in texts thanks to some useful functions like the possibility of creating a glossary. The GLOS course in video games localization has a module on CAT Tools.