LINGUIST List 4.986

Tue 23 Nov 1993

Qs: Hungarian, IRC, NLP, near

Editor for this issue: <>

Directory

1. , Query: Hungarian
2. Ken Miner, IRC
3. , 3rd Yr Proj in NLP
4. , Chest -> Near Again

Query: Hungarian noun inflection

Among Hungarian nouns with back vowel harmony, there are some which
take a low 'link vowel' (/a/) and some which take a mid link vowel
(/o/). Also, some nouns take /a/ as the 3 sg possessive suffix,
while others take /a/. Are these choices made entirely
independently or not?

In principle there could be four classes of nouns, so far as these
suffixal choices are concerned:
Class 1: link /a/ and possessive /a/
Class 2: link /a/ and possessive /ja/
Class 3: link /o/ and possessive /a/
Class 4: link /o/ and possessive /ja/.
Classes 1, 2 and 4 certainly have members, but I have the impression
that Class 3 may be unstable in that its members can also show up in
Class 4. Can any Hungarian expert shed light on this?

A similar question could be posed about front-vowel nouns in respect
of non-standard dialects which distinguish mid /e/ from low /E/ as
link vowels.

I believe that suffix choice interacts with stem allomorphy in
certain ways. If there are clearcut patterns I'd be glad to hear
about them, too.

Thanks in anticipation. I'll post a summary if there are
interesting replies.
Andrew
Andrew Carstairs-McCarthy
Department of Linguistics, University of Canterbury, Private Bag 4800,
Christchurch, New Zealand
Phone +64-3-364 2211; home phone +64-3-355 5108
Fax +64-3-364
 2065

Would anyone working on the pragmatics of IRC (Internet Relay Chat)
communications please get in touch with me? Bibliographic items, if
such exist, also welcome.

Ken Miner
minerkuhub.cc.ukans.edu

I am a final year Computer Science student undertaking a project in
natural language processing at the University of Leeds. My aim is to
construct frequency models of two- and three-letter combinations in
different languages using corpus material, so that the models could
be used to identify which language a piece of unknown text is in.

I am hoping to extend the project so that the model could be used to
disambiguate between a set of alternative letters in a word, obtained
from an OCR or handwriting recognition unit.

Please mail me if you know of any work in this area (useful texts,
references, or manufacturers of OCR/Handwriting recognition devices)
using linguistic models to improve recognition performance.

If you have any suggestions as to the usefulness of this project,
please be in touch.

Thanks,

Gavin Churcher.

John Limber has brought to my attention an 1885 publication by
Brinton of an even earlier ms. by Humboldt which says:

 In Maya TAN means in, toward, among; but it is also the breast or
front of the body...

I was wondering if any Mayanists out there could confirm or deny, or
emend this.