Bus announcements in Okinawa

« previous post | next post »

Travis Seifman noticed something interesting about the announcements on certain public bus lines in Okinawa: the pronunciation of Japanese / Okinawan place names in the English-language announcements is way off.

Although the quality of these two recordings may not be optimal, hopefully you can hear at the beginning of each recording the Japanese voice announcing the next bus stop, and then at the end of each recording, the English voice doing the same.  Travis says that he hears the announcer mispronounce Gi-bo as Gi-bah, and Kou-sei-en as Koh-sah-en.

The following paragraphs are Travis's ruminations on Gi-bah vs. Gi-bo and Koh-sah-en vs. Kou-sei-en.

There's actually something rather interesting going on here.  It's not so simple as just thinking, "Why did the Okinawan / Japanese staff not listen to the English recordings and realize the place names were getting so mangled?"  There seems to be an assumption that, since Japanese, and indeed many languages, alter foreign place names considerably, therefore, the Japanese / Okinawan staff don't need to review or judge the way the English-language announcer pronounces the place names. Maybe they're just assuming that Gibah is the English way of saying "Gibo" 儀保 and Koh-sah-en is the English way of saying "Kôseien" 厚生園, and assuming that this actually is the best way to communicate to English-speaking bus riders what the next bus stop is.

Given that the Japanese say Nyuu yooku ニューヨーク for New York, and Bankuubaa バンクーバー for Vancouver, and that English speakers say Spain, Germany, Rome, and Japan (instead of España, Deutschland, Roma, or Nihon), maybe it's only reasonable that the Japanese / Okinawan staff should listen to this English announcement and just assume that that's how English-speakers say it — rather than reviewing it and telling the woman she's done a poor job.

But, of course, that's not just how we say it. There is no standard English equivalent for Gibo or Kōseien. In truth, even allowing a little wiggle-room for pronouncing things just a little differently as feels natural to the English-speaking tongue (e.g., Toe-kee-yo instead of the Japanese way of pronouncing Tōkyō とうきょう), there is no reason that any of these place names should be pronounced all that differently from how they are in Japanese — indeed, without any standard English version of the place name, with this woman simply (mis)pronouncing each place name as she wishes, I can only assume the end result is something deeply misguiding and fairly useless for the rider who can't understand the Japanese announcements.

In the last three paragraphs of "Tones and the alphabet" (4/27/16), I describe how, on a train ride from Hualien to Taichung, Taiwan, the announcer read the names of the successive stations along the route in four languages:  Mandarin, Taiwanese, Japanese, and English, and how she systematically distorted the tones of the station names when she said them in English.  But supposing the announcer had inserted the correct Mandarin or Taiwanese tones in the English announcements.  How would that have sounded to monolingual English speakers who heard them?  The same questions hold for the English announcements of bus stops in Okinawa.  Maybe these assumptions about the way foreigners speak our language are well nigh universal.  Perhaps it would sound unnatural and even patronizing if we pronounced words from our language inserted in other languages exactly the way we do when we're speaking among ourselves.

Cf. "'Ni hao' for foreigners" (10/11/16).


  1. Ari Corcoran said,

    December 6, 2016 @ 10:08 am

    It's not just in Japan!

    It drives me mental to hear onboard train announcements in Italy insist I am travelling on Tren Itayliya, rather than Trenitalia. It makes no sense at all, given the English pronunciation is Itahly, and not Itayly. There is no possible rational explanation for this–unless someone can offer me one.

  2. V said,

    December 6, 2016 @ 11:10 am

    This phenomenon occurs in English-language transit announcements in many places throughout Japan, even in Tokyo.

    As Travis points out, what's curious about it is not that the "English" renditions are different from the Japanese – it's that they are also completely different from any conceivable way that a native English speaker would corrupt/mispronounce the Japanese, so one wonders *what process could have created these pronunciations*.

    I've never thought of a satisfying answer, but my top bets are that either text-to-speech or typos in the script handed to the English reader are involved.

  3. Keith said,

    December 6, 2016 @ 11:36 am

    The English language announcement sounds to me as if it is synthesized. Could it be that a computer has used rules of American phonology to create the recorded messages?

  4. Sergey said,

    December 6, 2016 @ 1:09 pm

    Speaking of Taichung, I've actually asked the locals, what is the right way to pronounce it – "Tai-jong" or "Tai-Chung", and they've told me "Tai-jong" in Chinese but "Tai-chung" is used in English. So I guess everyone just expects the English speakers to simply read the transliteration by the English rules.

  5. Jerry Friedman said,

    December 6, 2016 @ 4:46 pm

    When I'm trying to speak a foreign language and doing my best to get the sounds right, I use normal English for any English names that come up, but I occasionally wonder whether I should pronounce them the way a speaker of the language would. Of course, like the announcers described above, I might then get the pronunciations wrong.

    (When trying to speak Spanish here in New Mexico, I try to pronounce Spanish place names in Spanish—Santa Fe, Los Ángeles, etc. I've hardly spoken any French since college, and I don't think I've ever been faced with the question of how to pronounce "Detroit" or "St. Louis".)

  6. Simon P said,

    December 7, 2016 @ 6:03 am

    I always pronounce place names in the language I'm speaking. Pronouncing them in the local language would sound pedantic, unnatural and unhelpful. What you want is something that a native speaker would pair with the spelling in their language. Pronouncing Chinese place names in Mandarin in an English announcement would seem spectacularly unhelpful to me. The people listening in english would have a much harder time remembering the name of the place and wouldn't as easily be able to pair it with the note in the guidebook telling them what stop to get off.

  7. Axel said,

    December 7, 2016 @ 6:03 pm

    The English is definitely synthesized. And yes, probably American rules of phonology, or rather a more ore less sophisticated mapping of written English to the CMU phoneme inventory was used to create the input for the speech synthesis engine.
    The mapping for the bus stop names was added as a customization to a general English Speech-to-Text engine. In this case, the developer of the engine, the customizer, the worker typing in the custom pronunciation dictionary (with say, a couple of thousand entries), and the user (the bus company) are all different entities with very different levels of linguistic skill.
    To compound this problem, the worker making the dictionary probably has no way of running it through the STT to check the results, and the customizer and the user might miss the problematic words in their spot checks of the finished product (the English STT is just a small part of the system and no one has the budget to listen to all the entries),
    Never fear though.
    Development is rapid in the field and we'll surely have acceptable foreign pronunciation of Japanese proper nouns on busses in time for the Olympics.

  8. Joe said,

    December 8, 2016 @ 4:59 am

    If you go to 53 secs here you can hear how Shijo-Kawaramachi and Kawaramachi-Imadegawa is mangled by speech synthesis on a Kyoto bus:

  9. Victor Mair said,

    December 8, 2016 @ 7:41 am


    Thanks! That's great comparative material.

    This is what happens when a speech synthesizer trained to pronounce English uses its training to pronounce Japanese!

  10. mollymooly said,

    December 8, 2016 @ 8:44 am

    @Ari Corcoran If I saw the unfamiliar word "trenitalia" in an otherwise English-language text I would rhyme it with "genitalia". If the majority of those relying on the English-language announcements are native anglophones with zero knowledge of Italian, the anglicised novel pronunciation may be slightly easier for them.

    I agree that anyone with the slightest familiarity with Italian would find it less helpful; as would ESL listeners with no Italian.

  11. Joe said,

    December 8, 2016 @ 4:47 pm

    The speech synthesizer collapses with something like the name of the temple Chion'in. No example, sorry.

    Check out Kita-Shirakawa Shibusecho 北白川仕伏町 at 33 secs here:

  12. Dave Cragin said,

    December 8, 2016 @ 10:04 pm

    Speaking as a native English speaker, the most striking contrast to me is the English announcements on subways in Shanghai & Beijing versus those in New York and Washington.

    On these Chinese subways, the English is clear and easy to understand.

    In contrast, in New York and Washington, English is often spoken in a garbled manner that ranges from barely intelligible to unintelligible.

    I don’t really remember how the subway station names are said during the English announcements in China, but my vague memory is that they are said in the correct Chinese manner, i.e., with appropriate pronunciation/tones.

  13. Keith said,

    December 9, 2016 @ 9:17 am

    The discussion of speech synthesis awoke a memory of driving around California using a Tom Tom 920, set to speak English. There are two speech modes for that model of Tom Tom:
    – recorded human speech, which does not attempt to pronounce a word that has not been recorded, and
    – synthesized speech, which pronounces everything from stored text.

    Since California has so many Spanish place names, the difference between the words I read on signs and the words as prnounced by Tom Tom were sometimes very, very great.

    The one that comes to mind is "Sausalito", that Tom Tom rendered something like "Sɑːsɑːli:tɑː" where each syllable had a falling tone, making it sound vaguely Chinese.

  14. Rodger C said,

    December 10, 2016 @ 12:25 pm

    A totally techno-naïve question: How hard would it be, for heaven's sake, to produce an English speech synthesizer that didn't assume that, say, go has the same vowel as got?

RSS feed for comments on this post