Japanese issues with Siri / Shiri

I don't actually own a smartphone, but this video's been making the rounds since the introduction of Siri with the new iP***e. I thought I'd just make a small phonetics-related comment.

Listening to the speaker, the main problem here is his production of the NURSE vowel (typically /ɜ/ or /ɜr/),  which is a notoriously problematic one for many a non-native English speaker. The Japanese speaker produces the vowel closer to the THOUGHT vowel (typically /ɔ/). I'm just surprised that with a binary option: 'work' or 'home', Siri doesn't just go with what sounds closer to the 'work' option (but I'm not one who knows anything about programming, so I'll let the techsperts deal with that.)

I'm a little late to talk about this, but for the past 2 weeks, people have been talking about how siri means 'buttocks' in Japanese. Some people like this guy at TechnoBuffalo argue that it doesn't (it means nothing he claims). The thing is, technically, the 's' /s/ sound in Japanese never comes before the 'i' /i/ vowel: think of Japanese words that have been borrowed into English, like sushi, where 's' can precede the 'u' /ɯ/ vowel, but 'sh' /ʃ/ comes before 'i''.

However, in some transliterations of Japanese, you will see the word for 'I' written as both watashi and watasi. The reason is, as stated above, that the 's' sound can't come before 'i' and must be replaced by 'sh'. Therefore by default, the si in watasi will be read as shi. So siri by default, will be pronounced like shiri, which the internet would have me believe is Japanese for 'buttocks'.

Vowels: Does /j/ + /a/ = /æ/ ?

This is something that has been bugging me a little bit since I was in Nepal back in October last year. It concerns the orthographic representation of the English vowel /æ/ in words like taxi /tæksi/ when they are borrowed into Nepali.

The word for taxi in Nepali is ट्याक्सि, which transliterated gives Tyaaksi, where T represents a voiceless unaspirated retroflex stop (the tip of the tongue is slightly further back than when you produce a normal alveolar 't' sound in English). The appearance of the retroflex is not surprising here as English alveolar stops are usually borrowed into Indic languages like Nepali and Hindi as retroflex stops. (Assamese is the exception here as it has lost its retroflex stop series.)

The Nepali spelling suggests that the word is pronounced /ʈjaksi/ (/j/ represents the sound 'y'). My friend Sara insisted that Nepali speakers would palatalise (produce a particular speech sound while simultaneously raising the body of the tongue towards the roof of the mouth) the first consonant when saying the word and her theory was that they were trying to follow the American pronunciation of the word 'taxi'. (Correct me if I got this wrong Sara.)

I wasn't convinced with that explanation. For one thing, I would often hear speakers say something closer to /ʈɛksi/, without palatalising the first consonant and with the vowel /ɛ/, which is much more similar to /æ/. I also didn't see why Nepali would borrow English /t/ as a palatalised retroflex stop /ʈj/ since English /t/ isn't palatalised in this context. Also, I thought it counterintuitive that speakers would follow an American English pronunciation to guide their spelling, given the influence of British English across the Indian sub-continent.

Instead, my theory was that the combination of the letters 'y' and 'a' (या) represent the vowel /æ/ (or a close approximation like /ɛ/) and are not pronounced like /ja/. Unfortunately, other things came up (like 2 months in Nagaland), and I wasn't able to get more evidence to support my hypothesis.

Then today on the bus back to the Guwahati University Guest House, I found myself staring at the sign for the Volkswagen showroom while we were stuck in traffic. While I'm not as familiar with the Assamese script, the main thing I noticed was that the syllable corresponding to 'wa' in Volkswagen was written as ওয়া, which would be transliterated as 'oya'. Now, I'm assuming that the transliteration of Volkswagen in the Assamese script is based on the English pronunciation of the word and not the German one - the presence of 'o' in 'oya' suggests that it is trying to approximate /w/, not /v/. If this is true, then it provides evidence from another Indic language that a combination of the letters 'y' and 'a' are used to represent the English /æ/, as in 'wagon' /wægən/, or a close approximation of that sound.

I have to take a photo of that Volkswagen sign next time I pass by. Once I do and have evidence that Assamese does in fact use 'ya' to represent /æ/, I might try and confirm that Nepali uses the same strategy to represent this vowel.

(Note: standard Hindi avoids this problem altogether because it has the vowel /ɛ/ in its phoneme inventory, which is close to /æ/. It is often transliterated as 'ai' even though it is not a diphthong, e.g. in टैक्सी Taiksii 'taxi'. Also, the vowel in Bollywood actress Katrina Kaif's last name is not pronounced as a diphthong in standard Hindi.)

Inherent vowels in Nepali

People who are familiar with Indic writing systems will know that unlike in alphabetic writing systems (like the Latin alphabet we use in English), consonants in such scripts have an inherent vowel associated with them, e.g. in Hindi, क represents 'ka' (pronounced more like [kə] with a schwa), while कि represents 'ki' [kɪ] and के 'ke' [ke], with the vowel symbol replacing the inherent vowel of क. (It is is this replacing of the inherent vowel that separates such scripts from syllabaries such as Japanese hiragana.)


Having learnt some Hindi before attempting Nepali, I started by assuming the inherent vowel in Nepal was also pronounced as schwa [ə]. I quickly learnt from Lauren and from my own ears that it is not a schwa, but rather an open back vowel, closer to [ʌ] (like the vowel in 'hut'). This analysis was further substantiated by Khatiwada's 2009 phonological description of Nepali as part of the Journal of the International Phonetic Association (JIPA)'s Illustrations of the IPA series which gives /ʌ/ as the underlying vowel phoneme. (Incidentally, I'm waiting to hear back from them regarding an article draft I submitted a few months ago.) However, Lauren and I also noticed that this vowel could be pronounced with rounded lips [ɔ] (like the vowel in 'hot'). Khatiwada notes such variation for this particular vowel, along with other variants, depending on the speaker and surrounding consonants (basically, there's a whole lot of factors and it's too hard to tease out right now.)


It was therefore with some satisfaction that my Nepali tutor asked me yesterday to say 'dog', which I pronounce [dɔg], and 'duck' [dʌk] (the latter with an unreleased [k], but the vowels themselves distinctly different to me). She then asked me if they were pronounced the same, providing some evidence that to her ear, [ɔ] are [ʌ] are the same sound (or, allophones of the same phoneme).

Of course, for an English speaker such as myself, it's still quite annoying because since I hear the difference between the two vowels all the time, sometimes I just want to know which one I should use!