Wednesday, 19 August 2009

chequered speech-to-text

No time for a proper blog entry today, so here’s a silly piece of verse you can find in various versions on various sites around the internet.
Eye have a spelling chequer,
It came with my Pea Sea.
It plane lee marks four my revue
Miss Steaks I can knot sea.
Eye strike the quays and type a whirred
And weight four it two say
Weather eye am write oar wrong
It tells me straight a weigh.
Eye ran this poem threw it,
Your shore real glad two no.
Its vary polished in its weigh.
My chequer tolled me sew.
A chequer is a bless id thing,
It freeze yew lodes of thyme.
It helps me right all stiles of righting,
And aides me when eye rime.
Each frays come posed up on my screen
Eye trussed too bee a joule.
The chequer pours o'er every word
Two cheque sum spelling rule.

Most spell checkers will find no mistakes here, because they check words in isolation. We can see the mistakes, because we can consider the context in which each word occurs. That’s why homophones are such a problem for automatic speech recognition (speech-to-text).
Some versions start not “Eye have” but “Eye halve”. That’s fine for Americans but doesn’t work for people like me who make a distinction between hæv and hɑːv.
I suspect that Americans don’t actually use the spellings cheque and chequer at all.
But there is one word even in this version which does not work for me. Can you see which it is? (And I’m not talking about jewel, nor about dodgy weak vowels.)

21 comments:

  1. Is it "vary" instead of "very"? I think "vary" is /veri/ in British English, while "very" is /vɛri/. (I'm American, so I don't know for sure).

    ReplyDelete
  2. It’s that you pronounce “whether” with unvoiced [ʍ], right?

    ReplyDelete
  3. ascent1729 is right, but for the wrong reason. The FACE vowel never occurs before /r/ in RP and similar accents (except across a morpheme boundary), but rather /eə/ (aka /ɛː/). So "vary" has the SQUARE vowel, but "very" has the DRESS vowel. Everyone outside North America has a three-way distinction "merry-marry-Mary".

    ReplyDelete
  4. I'd add that "Miss Steaks" is not really a homophone of "mistakes".

    ReplyDelete
  5. Decades ago, American bank employees used cheque internally, to distinguish it from check 'verification'. But that's been gone for a long time.

    ReplyDelete
  6. Here are a set of possible non-homophones.

    plane lee / plainly = FLEECE / happY

    four / for = FORCE / NORTH

    Your / you're = FORCE / CURE for some

    shore /sure = FORCE / CURE

    vary / very = SQUARE / DRESS (distinct for all non-North American accents)

    come posed / composed = STRUT / schwa (distinct for me)


    miss stakes / mistakes could be different for those with the weak vowel merger if "miss" is stressed

    bless id / blessed ditto

    whirred / word
    weather / whether

    ReplyDelete
  7. It really depends on the accent.

    - eye/I. eye gets more stress.

    - plainly/plane lee. Only with HAPPY tensing, so happy the close i is long or diphthongal. (Occurs, of course.)

    - for/four. Not just the diphthong, but even in today's EE/RP, the former is /fə(r)/ here, not /fɔ(r)/.

    - or/oar. Diphthong.

    - blessed/bless id doesn't work for many.

    - sure/shore. Different diphthongs.

    - to/two. A classic, and I never liked it, though I giggle about this. An American friend of mine claims he pronounces them the same, not /tə/ vs /tuː/.

    - very/vary.

    - you/yew. /jʊ/ and /juː/ here.

    - composed/come posed. Usually /kəm/ and /kʌm/.

    - to/too.

    - jewel/joule.

    - cannot/can knot and mistakes/Miss Steaks are disputable concerning the consonatant length.

    ReplyDelete
  8. Actually, all of those pairs are distinct for my speech except the last two and your/you're. I suppose that for/four are distinct for most people in that "for" has a weak form whereas "four" doesn't.

    come posed/composed are so different for me that I didn't even realised what the poem was trying to say there. In "composed", I would use the LOT vowel for the com- prefix; Jack Windsor Lewis has observed that Northern English people of all social classes do this.

    ReplyDelete
  9. What vowel do Northeners have in come?

    ReplyDelete
  10. Whirred for word and weather for whether don't really work for me. I have [ʍ] fairly regularly, when stressed at least. (I don't think I'd say "wirred" though, ever.)

    ReplyDelete
  11. What about "shore"? I doubt if JW has that as a homophone of "sure". And I for one can't picture an accent where "come posed" works for "composed". Also, those dodgy weak vowels do spoil things. You have to imagine the piece read by a non-native speaker who has not learned to obscure the vowels of words like "to" in a way that doesn't happen with "two".

    ReplyDelete
  12. And I for one can't picture an accent where "come posed" works for "composed".

    I think this could work in GenAm, provided that "come" is unstressed.

    ReplyDelete
  13. I'm like Michael - I distinguish /w/ and /hw/, so "whirred" for "word" and "weather" for "whether" fail for me.

    ReplyDelete
  14. "to" is a homophone of "two" when stressed, as in "Whom did you speak to?"; similarly, in NORTH-FORCE merging accents stressed "for" is a homophone of "four". I guess that's what Wells meant by "And I’m not talking about ... dodgy weak vowels."

    The obvious answer would be "your shore" for "you're sure", but he said "one word" and these are two, so I guess he merges them. So I think ascent1729 got the right one.

    (BTW I distinguish "joule" from "jewel" for a different reason, as I pronounce the unit of energy rhyming with "owl". As the father of the namesake of that unit advertised his beer as, "Do you pronounce it Joule's to rhyme with schools, Joule's to rhyme with Bowls, or Joule's to rhyme with Scowls? Whatever you call it, by Joule's or Joule's or Joule's, its GOOD!", I guess I am allowed to pronounce it whichever way I like.)

    ReplyDelete
  15. "to" is a homophone of "two" when stressed, as in "Whom did you speak to?"

    No, that would at most be /tʊ/, not /tuː/ as in the awkward but possible question "Whom did you speak, too?". (Is that really identical in GenAm, professor Wells? Or even in cispondian English, and I'm only imagining things?) Apart from that, the text above doesn't suggest stress.

    Joule seems to have pronounced his name with a /nuː/.

    I'm afraid I don't understand what the difference between your and you're might be, whether they are mono- or diphthongal.

    ReplyDelete
  16. Lipman, I suppose that like some people say "we're" as if it was "we are" (instead of "where"), some people also say "you're" as if it's "you are" (instead of "your"). I've heard it once or twice, but it sounds more like "you were" to me.

    And I have /tuː/ as the final form of "to", not /tʊ/. The vowel is phonetically short, but the quality is unmistakably that of /uː/ being fronted (i.e. approximately [tʉ]). When you just start learning to read, you don't normally reduce words because you can't read fast enough (and you haven't got used to extending your view of the text towards the right), so it's common to learn that "to" is pronounced as /tuː/. This may not be the case for you-personally, so before telling people off for reporting how they speak, I suggest you realise other people probably speak differently.

    (As for sure vs shore, I was taught that these were homophones in school, and when I type I often start "short" as "sur" before I realise I have no way of continuing.)

    ReplyDelete
  17. Pressing the "Preview" button ended up publishing that comment before I'd intended. I would've revised the ending of the middle paragraph before I'd sent it if I had've had a chance to reread it, so I apologise for the tone. Why is it so hard to post comments on this blog?

    "the quality is unmistakably that of /uː/ being fronted"

    "being fronted" here should be read as "because it's fronted", i.e. the subject of "being" is "the quality" and not "/uː/" as it seems to read.

    ReplyDelete
  18. I'm sorry if I hurt your feelings, though I'm not sure how I did that.

    Otherwise, I'm afraid I don't understand what you mean, and where I have a hunch, I think you're mistaken. Also, that pupils read each word or each syllable "unreduced" is hardly relevant for normal speech or for the spell checker fun, let alone what that gym teacher who's also giving the English lessons tells them.

    ReplyDelete
  19. @ Lipman: most Northern English people have /ʊ/ in "come", so /kʊm/, but a schwa can be heard in unstressed positions /kəm/.

    ReplyDelete
  20. Fernando Lamadrid23 January 2010 at 21:15

    I make thr Mary, merry, marry distinction

    ReplyDelete