Published: April 25, 2022

Author: Madelyn Weber
Nominator: Orin Hargraves
Course: LING 3430 Semantics, Fall 2021
LURA 2022

As languages grow further apart in terms of their similarities, translator systems often become less accurate in producing accurate and grammatical translations. An example of this is in regards to the use of Keigo - honorific language - in Japanese. In a broad sense, Keigo is a way of marking language in different levels of politeness depending on who is speaking to whom and in what context. Overall, Japanese is a highly context-dependent language, and that is reflected in this system of polite speech. Japanese grammar marks different levels of honorific language by vocabulary and grammar - something which English does not do. 

We can think of Keigo as having three different “levels”: 

Casual Speech - used between friends, romantic partners, or toward those who hold a lower social status from the speaker 

Polite Speech - used in most day-to-day interactions, such as when communicating with strangers, or, for example, when a shopper is addressing a store employee 

Honorific/Humble Speech - often used in the same situations, and are dependent on who the subject of an utterance is. If the speaker is addressing their boss, company CEO, professor, or even the elderly, they would opt for honorific speech when speaking about the listener, but would use humble speech when speaking of themselves, which essentially adds an extra layer of “distance” between the speaker’s and the listener’s difference in social position 

This system can present problems for online translator systems due to how highly context-dependent it is; especially when translating from a language that holds no such notion of formality embedded within its grammar, such as English. 

Let’s take a look at some ways in which discrepancies in the usages of Keigo produced by Google Translate occur when using English → Japanese directed translations. 

Translation 1: Asking a simple (polite) question in Japanese.

Sentence breakdown: 

なんて [nante] - what

言い [ii] - say/ask 

ました [mashita] - (polite, past-tense form) 

か [ka] - (question marker) 

Translation 2: Putting output from translation 1 back into the translator.

Sentence breakdown: 

何って [nantte] - what 

言 [i] - say 

った [tta] - (casual, past-tense form) 

の [no] - (casual, question marker) 

As can be seen from translations 1 and 2, we are getting two different Japanese grammatical forms, one casual and one formal, from the same English sentence. As English has no means of encoding the level or respect one desires to use from its grammar alone, we have no idea as to what level of honorific language the translator will produce for the Japanese translation. We can look into one more example to further exemplify such discrepancies. 

Translation 3: Using the same English sentence, but with a different pronoun.

Sentence breakdown: 

彼ら [karera] - they 

は [wa] - (grammatical particle) 

何 [nan] - what 

と [to] - (grammatical particle) 

言い [ii] - say 

ました [mashita] - (polite, past-tense form) 

か [ka] - (question marker) 

Notice how we switched up the pronoun in the English sentence from “you” to “they”. From this alone, we are now getting a formal Japanese output. This exemplifies the discrepancies in the translator system’s ability to form sentences; sometimes they come out very casual and other times, formal. In English, there is no such notion of the pronoun “they” being more formal than “you”, and this holds true in Japanese as well, thus offering no clear explanation as to why the addition of a different English pronoun would provide us with a different formality in the Japanese output. 

The translation examples above are just a snapshot as to the discrepancies that online translators exhibit when it comes to producing Japanese translations and the usage of honorific language within them. Such discrepancies can become a problem for someone who wishes to use an online translator system if they do not speak Japanese, or understand the cultural significance of using the correct forms of Keigo. For this reason, there are two proposals that I would make to potentially increase the correctness of such online translator systems, first, to always output sentences in the polite form, as this is considered to be the general form that would be used in most day-to-day interactions, and is the best form to fall back on to avoid accidentally offending someone. My second proposal would be to add a means of marking which “level” of politeness one wishes their sentence to be output as. Although it wouldn’t be perfect, it would help hinder any obstacles in communication one may face when using an unpredictable translator system.

Header image credit: