Allegro Rules
Capita Selecta Language Processing
Instructors: John Nerbonne, Gertjan van Noord, Wilbert Heeringa
(course under development)
Course Number: LIX010B05
Topics for Bachelor Theses
Please see below.
Course Description
The goal of this project-oriented course will be to implement some
"allegro rules" (also known as "fast speech rules"
or Schnellsprechregeln) in phonology and then to compare
dialect pronunciations on the one hand with standard speech when it is
produced quickly on the other. Schmidt and Herrgen (1989) have
conjectured that measures of "dialectality"
(Dialektizität) might be infected by recordings of
standard speech as it is modified by fast speech rules.
Herrgen-Schmidt
Herrgen and Schmidt (1989) have conjectured that measures of
dialectality (Dialektizität) might be infected by the fact
that the transcriptions (or recordings) of dialectal speech reflect
the use of (standard) allegro rules. If these are compared to
standard speech in which allegro rules have not applied, then the
measurements are misleading. That is, some dialect recordings,
including transcriptions, appear to be nonstandard, but only because
the recording (transcription) shows an allegro form, rather than a
so-called "citation-form". In this project, we will attempt
to develop software to investigate how large these effects are.
What are "allegro rules"?
Fast speech rules in German include rules such as post-tonic schwa
deletion ([hAp] for habe, [wErdn] for werden, [nOIn] for
neuen as well as neun, but also the simplification that
leads mehr and Mäher to be pronunced the same); the
syllabification of nasals ([wErd=n] for werden); nasal
assimilation to preceding consonants ([hAb=m] for haben,
[bIt=n] for bitten, [hak=N] for hacken, even [lif=F] for
liefen); the deletion of obstruents before syllabic nasals
([hAm] for haben, [lem] for leben, [reN] for
regen); epenthesis of homorganic stops ([kOmpst] for
kommst, [rEntst] for rennst, and [zINkt] for
singt); palatalisation of /s/ before palatals ([aiSraNk] for
Eisschrank); degmination, i.e., deletion of homorganic
obstruents ([EN=lain] with a single [l] for Engellein) (see
Wiese, 2000, p.231ff); elimination of initial glottal stops ([AbEnt]
instead of [?AbEnt] for Abend), vocalizing final /r/ ([Aba] for
aber), or simplifying the affricate /pf/ to [f] ([fEniC] for
Pfennig). One task will be to find a list of allegro rules,
ideally all those that play a role in the sample of data used (see
below).
Note that other obligatory rules may be used together with allegro
rules, such as final devoicing ([hAp], not [hAb] for habe).
It is not intended that we incorporate the effect of
allegro rules if these are used regionally, e.g., the deletion of
final /t/ in words such as nicht, vielleicht or sonst,
which is found in northern and middle Germany, but not in the south.
(Ramers & Vater, 1995, p.50).
Implementing Phonological Rules
Finite-State
Automata Utilities (also known as "FSA Utilities") is a
Prolog package which facilitate the implementation of phonological
rules. Note the tutorial for FSA Utilities.
Calculating Pronunciation Distance
L04 is a software
package designed for the measurement of pronunciation differences.
For our purposes it will be important to use the software to find the
least distance between a dialect recording and the whole set of possible
allegro rule outputs. Note the tutorial on the use of L04.
Data
The data will we use to test the work and also to test the idea that
allegro rules are infecting the observation of dialectality is the
data of the Phonetischer Atlas Deutschlands (PAD), which we
have digitized in Groningen. The data is available here,
transcribed using the system of X-SAMPA.
Schedule
Week |
Date |
Theme |
Leader |
Material |
1 |
6 Feb. |
Intro |
Nerbonne |
Background
|
|
|
|
|
Pronunciation Comparison
|
2 |
13 Feb. |
Phonetic Transcription |
Nerbonne |
Phonetics
|
|
|
Implementing Phonology |
Van Noord |
FSA Basics |
3 |
20 Feb |
Phonology |
Nerbonne |
Phonology
|
|
|
Implementing Phonology |
Van Noord |
Ex.1, Ex.2 |
Lab! |
22 Feb! |
L04 |
Heeringa |
Tutorial |
4 |
27 Feb. |
Project Planning |
Nerbonne |
|
Lab! |
1. Mar |
FSA |
Van Noord |
Tutorial |
5 |
6 Mar. |
Project Meeting |
|
|
6 |
13 Mar. |
Project Meeting |
|
|
7 |
20 Mar. |
Final Meeting |
|
|
- (Recommended) Pick out one subtopic from those in the project
above and work on it in more detail, e.g.
- seek a more complete set of allegro rules and implement them
and evaluate their effect;
- work out a what a null hypothesis should be in a test of
whether the allegro rules differ significantly, attempting to
allow for the fact that by taking a minimum difference from
one set to another, one is guaranteed to find a difference no
larger than the first
- (In combination with a course on Geographic Information Systems):
Explore the Monmonier algorithm as a means of identifying significant
dialect boundaries. See Manni, Heeringa and Nerbonne (2006)
Are Family Names just Words? Comparing Geographic Patterns of
Surnames and Dialect Variation in the Netherlands. Accepted to
appear in Literary and Linguistic Computing 21(4), 2006.
- (In combination with a course on Geographic Information Systems):
Implement a GIS which can calculate travel times in the 19th century
in Germany. See van Gemert (2002)
Het geografisch verklaren van dialectafstanden met een GIS.
Literature
Wolfgang Dressler (1975) "Methodisches zu Allegro-Regeln."
In: Wolfgang Dressler & F.V.Mares (eds.) Phonologica 1972
Munich: Fink. pp.219-234.
Herrgen, Joachim & Jürgen Erich Schmidt (1989)
"Kontrastive Dialektgeographie" in Wolfgang Putsche, Werner
Veith, & Peter Wiesinger Dialektgeographie und
Dialektologie, (Deutsche Dialektgeographie (90)),
Marburg: N.G.Elwert, pp.304--346.
Ramers, Karl-Heinz & Heinz Vater (1995) Einführung in
die Phonologie Hürth: Gabel.
Wiese, Richard. (2000) The Phonology of German Oxford:OUP.
John Nerbonne
Last modified: Tue Apr 4 14:52:34 CEST 2006