Route: spelling and pronunciation [and IPA vs. Orthographic respellings]

[Mark A Mandel]

On Tue, 2 Jul 2002, Thomas Paikeday wrote:

#This is what I call taking the bull by the horns. My answers are
#interspersed below right after the questions. Thanks for asking.

Thanks for the explanation. I now understand much better the context of
this system.

#On the defensive, I'd say the system is tailored to the needs of the
#popular user, not for academics.

Clear enough, and agreed. -- Without quoting or remarking on your whole
post, I'll agree that it's appropriate to that purpose, although
definitely not for describing pronunciations on this list.

        [Mark M.]
#> I had to face a very similar issue in constructing representations of
#> pronunciations at Dragon Systems. A senior scientist kept insisting that
#> there had to be some way of spelling each phoneme distinctly and
#> transparently to the naive user, answering each of my counterexamples
#> simply by shaking his head and repeating, "There's GOT to be a way."
#> There isn't: SOME learning is required, e.g., OOH for [u:] (the "oo" of
#> "mood" to you, Tom, or the "ui" of "suit") vs. OO for [U] (the "oo" of
#> "foot"). And it's especially impossible (imho) if you don't allow the
#> option, as we did, of separating multi-letter phoneme symbols with
#> spaces or hyphens: that's where the ambiguity of "PYE" arises.
#
#ANSWER #2: Mark, I feel for you and almost completely agree with you.
#About the learning part, I believe the average elementary school
#graduate should be able to read (OOH) as it is meant to be read, not in
#isolation, but in well-known phonetic contexts. When you say "OOH for
#[u:]" that looks like a system for system's sake as in the Oxford
#American Dictionary and others, and that means more learning.

For a dictionary, "system for system's sake" may be an appropriate
dismissal. For a speech recognition company, though, system is
absolutely essential. We *had* to provide for users to say "Pierre" and
"Khrushchev"  (as both /'krus.CEv/ and /'kruS.CEv/, and maybe /kruS.COf/
as well -- C and S are c/s-hacek, E and O low-mid), and a whole slew of
other non-English words and, especially, names.

#We use a shoehorn to distinguish between "long" and "short" in ambiguous
#phonetic contexts. Thus for "foot" (granted the vowel could be either
#long or short) we give the pronunciation as (short "oo"). For "food," we
#say (long "oo") and for "good" (short "oo").

Ah. This "shoehorn", as exemplified below in your reply, is what I
didn't recall seeing in the earlier discussion. Extranotational
annotation, like mine above for "Khrushchev", can cover anything. In
this case, you rely on it (appropriately, imho) to cover distinctions
not made by the respelling.

#We also don't bother about the nicer phonetic questions involved,
#like whether the long "oo" is diphthongized or not, just as no
#indication is given for aspirated/unaspirated consonants as in
#"pin/spin" (a question Rudy raised) in the best of IPA and diacritic
#systems used in dictionaries.

Agreed, these would be wholly excessive for your purpose.

#> Come to think of it, how do you respell "suit" and "foot"?
#>
#ANSWER #3: "foot" is respelled as explained above, but if anyone wants
#to check whether it's OK to say (SYOOT), that is not within the scope of
#a midsize dictionary.

My bad example. I meant the distinction between the nuclei, which is
made by your annotations:
        suit    SOOT (with long "oo")
        soot    SOOT (with short "oo")
 -- yes?

#> BTW, did you ever answer my other question, about dialectal
#> neutralizations such as which/witch and cot/caught?
#
#ANSWER #4: We leave this to the "tutored" user. This applies to all
#wha-/whe-/whi- words. Similarly, users may go for what is usual in their
#own dialects on the cot/caught question. Normally (cot) should suffice
#for both words. But we do use the "aw-" respelling on occasion.

Now, here's a problem. Discussion below, after the quote of your summary
of principle.

#We try to be user-friendly (not scholastic) and do the best job we can
#with the best tools at our command. I appreciate scholars raising
#academic questions because the system has to make linguistic sense (I
#hope mine does), but most users, we believe, couldn't care less (no
#offence!) about systems and keys to systems and if they did, they could
#go to a bigger dictionary like Webster's Third or OED.

I understand your argument, and I suppose you're right about "most
users". But presumably the reason your users consult the pronunciations
is that they don't know how to pronounce a particular word. Your use of
"normally" suggests that the cot/caught distinction is exceptional, so
far outside the (descriptive) norm that you're justified in generally
dismissing it. If a user speaks a dialect that maintains the
distinction, how is she supposed to know from your merged respelling
which vowel to use in an unfamiliar word?

To take a more plausible example than "cot" and "caught", imagine a New
York City high school junior looking up the unfamiliar words "sot" and
"fraught" (more likely to be unfamiliar than the minimal example
"sought"). The "o" respelling will tell her, misleadingly, that she
should rhyme them. How is she to know otherwise?

-- Mark A. Mandel
   Linguist at Large