Machine Translation (MT) between Japanese and English

MT (Machine Translation) is the single biggest buzzword of the industry in recent years, and with good reason. Intelligent use of MT can be an effective tool in your localization strategy kit. But let’s emphasize that only intelligent use makes sense: applying MT without understanding its limitations, without knowing your own quality requirements and without understanding the level of post-editing required, is a recipe for disaster. As an example of pitfalls of which you should be aware, today I will look at the example of Japanese  English machine translation and examine some of the differences….and dangers.

(1) There are huge differences in grammar and semantics between English (and Indo-European languages generally) and the Japanese language.

Besides very different grammar rules, Japanese also has very different syntax, word order, etc. There are also differences in the ending elements – inflections in the form of bound morphemes if you want to get technical about it – that are driven not just by meaning and grammar but also by differing levels of formality, e.g. normal, polite or very polite form. And depending on context, the consistency of those ending elements may need to be kept for uniformity. In other words, the Japanese language has a greater number of variations in sentence construction and requirements than does English. When using MT it can therefore be extremely difficult to match appropriate translations between these two languages.

(2) Modern Japanese sentences can utilize as many as three writing scripts.

Japanese has three distinct scripts: Hiragana (a syllabary used for grammatical elements and certain verbs and verbal forms), Katakana (a syllabary used for foreign terms that have no equivalency in Japanese) and Kanji (adoptive Chinese characters used for subjects, verbs, adjectives, etc.).

But aren’t the rules pretty well defined for which one is used where, meaning shouldn’t MT be able to resolve this? Not necessarily. Depending on the target audience in the Japan market (or the client reviewers’ demographical background), it can vary.

(3) The unique competitive environment of the Japanese market

The final reason for caution when using Japanese  English MT isn’t linguistic or technical, but cultural and commercial. Japan is the world’s third largest economy, and as such represents a huge market. Japan therefore has many domestic products with which you will be competing. Besides having a ‘home field advantage’, your local competition is of course writing their materials in Japanese, while foreign competitors have to translate. When foreign competitors don’t choose the right localization partner, their materials end up sounding foreign, unnatural. Using MT only increases this risk and disadvantage.

Conclusion:

MT is tricky at the best of times, even between similar languages. When it comes to using MT between Japanese and English, clients should not just sign up for MT + post-editing services from their localization vendors unless they themselves have someone in-house or a trusted third party to review the MT + post-editing test result and evaluate how good the MT engine really is. If the MT is really up to snuff, then post-editing will be true post-editing, not rewriting disguised as editing, which can end up adding more costs and time than it saves.