MLU in syllables?

Tue Sep 10 23:49:46 UTC 2013

Actually,  you don't even need
a good correlation. All you need is to identify a function relating mlu in
syllables to mlum that has a good enough fit. Then you can always predict
with a certain degree of accuracy mlum from mlus.
Bruno Estigarribia
UNC Chapel Hill
Hi Yvan and all,

Thanks for your response and willingness to consider implementing something
like this in Phon.

You're right that morpheme-based MLU makes most sense for Inuktitut, and
I've already done that for my own child language corpus. But, as you point
out, this involves first analyzing utterances morphologically - not a
trivial task given the morphological complexity of Inuktitut. This would be
fine for people with training in Inuktitut linguistics. But the people
doing the language assessments are typically pediatricians or SLPs who
speak little Inuktitut and rely on parental report, or native speaker
medical assistants who do not have the requisite linguistic training (and
anyway wouldn't have the time for morphological analysis). Thus, we need
another solution.

Determining the number of syllables in an utterance is very doable for
native speakers with no linguistic training, because Inuktitut is written
in syllabic characters. So my plan is to calculate both the MLU in
morphemes and the MLU in syllables on the data I already have, to see
whether there is a high enough correlation between them. If the correlation
is high enough, then it would be reasonable to use MLU in syllables as an
indicator of linguistic ability for Inuktitut.

The data I have for now are already analyzed morphologically and I've
calculated MLUs using CLAN. We've also figured out a way to do syllable
counts that's workable for now. But if this idea works out and the
correlation is high enough, we may well want to test it on more data that
isn't coded. And then I'll definitely be in touch!

My request to info-childes was more to figure out if anyone has done
something like calculating MLU in syllables before. If so, we could benefit
from their experience and not reinvent the wheel. In this regard, thanks to
Nan and Sarita for the references - very helpful.

Best,
Shanley.

On Sep 10, 2013, at 1:11 PM, Yvan Rose wrote:

Dear Shanley, everyone,
MLU counts by syllables would potentially work for Inuktitut. However, I
wonder whether a morpheme count wouldn't work even better, closer to the
original spirit of MLUs, as the syllable count would elude anything
meaning-related.

Building on Nan's suggestion, I confirm that we could easily rig both a
syllabification algorithm (for Inuktitut), and a (general-purpose)
syllable-based MLU count in Phon. I would be happy to work with you towards
both objectives. This also raises the question of converting your data into
the Phon format, something that shouldn't pose any serious problem either
given our advances on this front over the last year.

A morpheme-based one would require the utterances to be morphologically
analyzed first. We did this for Cree a few years ago. Attached is a screen
shot illustrating this. Please feel free to contact me separately to
discuss the specifics.

Best regards,
Yvan

<Morpheme-basedMLU.png>

On 2013-09-09, at 22:12, Nan Bernstein Ratner <nratner at umd.edu> wrote:

OK, then, pragmatic issues satisfied, Shelley Brundage and I tried this a
while back for English and came up with a conversion factor of about 1.5
for American English speaking kids’ speech from about 3-8 (JFD, 1989). But
you are asking about norms? Because you seem to have identified that the
conversion is closer to 1:1 for Inuktitut. This seems to be a question Phon
users might be able to compute for longitudinal samples in the database. I
guess Brian could say whether any dataset in TalkBank is pre-syllabified.
If so, one could theoretically do this on archival data.****

Best regards to all,****
Nan****

*From:* info-childes at googlegroups.com [mailto:info-childes at googlegroups.com]
 *On Behalf Of *Shanley Allen
*Sent:* Monday, September 09, 2013 6:04 PM
*To:* info-childes at googlegroups.com
*Subject:* Re: MLU in syllables?****
** **
No, it's a good question. I wouldn't try this for English, but it makes
sense for Inuktitut because Inuktitut is written with a syllabic script. So
anyone who writes Inuktitut (98% literacy rate!) is forced to think in
syllables, and native speakers of Inuktitut who are assisting in the
assessment can easily transcribe the child's speech in syllabics and then
simply count the number of characters in each utterance. Those same native
speakers would have a very hard time dividing Inuktitut into morphemes to
calculate a traditional MLU. And assessing MLU in words (as is done for
Irish, for example) is uninformative for Inuktitut because it is
polysynthetic and has rampant argument ellipsis so all the interesting
linguistic development at early ages happens in the morphology.****
Shanley.****
** **
** **
On Sep 9, 2013, at 11:48 PM, Nan Bernstein Ratner wrote:****

****

This may seem like a stupid question, but do pediatricians know what a
syllable is? My spouse, who is a doctor, probably would prefer something
based on a literate unit, like words. I am afraid to ask what he thinks
about counting syllables in any language, but I will go home tonight and
ask.
Nan Bernstein Ratner

-----Original Message-----
From: info-childes at googlegroups.com
[mailto:info-childes at googlegroups.com<info-childes at googlegroups.com>]
On Behalf Of Shanley Allen
Sent: Monday, September 09, 2013 5:25 PM
To: info-childes at googlegroups.com
Subject: MLU in syllables?

I'm wondering if anyone has tried to calculate MLU in syllables for any
language, or knows of any study like this. I'm working with an SLP
colleague to devise tests that would be suitable for assessing language
delay in Inuktitut. The target user group is pediatricians who don't know
Inuktitut well enough to do a morphological analysis, but want to use
measures that are appropriate for the language. This idea seems a little
crazy, but our preliminary results show that MLU in syllables correlates
very highly with MLU in morphemes in Inuktitut data, so it might just work.

Thanks in advance,
Shanley Allen.****

Shanley Allen
Professor, Psycholinguistics and Language Development
Dept. of Social Sciences, University of Kaiserslautern****

 --
You received this message because you are subscribed to the Google Groups
"Info-CHILDES" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to info-childes+unsubscribe at googlegroups.com.
To post to this group, send email to info-childes at googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/info-childes/A653E233-5C6B-42A5-A541-36C9DEC53680%40bu.edu
.
For more options, visit https://groups.google.com/groups/opt_out.

-- 
You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
To post to this group, send email to info-childes at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/CAMWGojmQRO%3D41-zoo8ZtnpFKVEb7d-%2BWXqYX9mp4dSFfZa%3DYgw%40mail.gmail.com.
For more options, visit https://groups.google.com/groups/opt_out.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/info-childes/attachments/20130910/f7fa6bae/attachment.htm>