Conference PaperPDF Available

Spatial Variation of Articulation Rate and Phonetic Reduction in Standard-Intended German

Authors:

Abstract

Tempo of speech and phonetic reduction are closely related and differ in their spatial distributions. The SpuRD-project (Sprechtempo und Reduktion im Deutschen) focusses on this web of relationships and their spatial variation for the whole German-speaking area in central Europe. Using standard-intended reading material in normal and fast reading tempi, an array of reduction phenomena is analysed to find whether they are caused by tempo or whether they are independent variants of a limited linguistic area. The results reported here show on a macroscopic level that the spatial distributions of temporal characteristics such as the duration of articulation and the degree of segment reduction do not coincide everywhere, but have independent distribution areas especially at a higher tempo. That means that articulation rate is composed of regionally varying temporal and segmental features. For explanations of this macroscopic variation the material is analysed with regard to particular microscopic variation with independent spatial distributions.
SPATIAL VARIATION OF ARTICULATION RATE AND PHONETIC
REDUCTION IN STANDARD-INTENDED GERMAN
Matthias Hahn and Beat Siebenhaar
Universität Leipzig, Germany
matthias.hahn@uni-leipzig.de, siebenhaar@uni-leipzig.de
ABSTRACT
Tempo of speech and phonetic reduction are closely
related and differ in their spatial distributions. The
SpuRD-project (Sprechtempo und Reduktion im
Deutschen) focusses on this web of relationships and
their spatial variation for the whole German-speaking
area in central Europe. Using standard-intended
reading material in normal and fast reading tempi, an
array of reduction phenomena is analysed to find
whether they are caused by tempo or whether they are
independent variants of a limited linguistic area. The
results reported here show on a macroscopic level that
the spatial distributions of temporal characteristics
such as the duration of articulation and the degree of
segment reduction do not coincide everywhere, but
have independent distribution areas especially at a
higher tempo. That means that articulation rate is
composed of regionally varying temporal and
segmental features. For explanations of this
macroscopic variation the material is analysed with
regard to particular microscopic variation with
independent spatial distributions.
Keywords: speech tempo, reduction, geophonetics,
regional variation
1. INTRODUCTION
Speaking ‘standard’ means, at least for German,
speaking without using salient regional pronunciation
features [15, 11]. Most speakers are likely to use
regionally accepted variants of the standard, some of
which are well documented [1, 6]. However, regional
prosody within the standard has not yet been
investigated. There is much work concerning reduced
forms, especially so called weak forms [14, 22].
However, these studies are mostly based on a corpus
labelled as “standard material”, which in fact contains
mainly material from standard speakers of the
northern German-speaking areas. Admittedly, there is
a certain difference between “best standard in use”
and its assessment [11]: the northern variant normally
is assessed as the better standard. Nevertheless,
variants of the middle and southern German area as
well as those of Austria, Switzerland, Luxembourg,
East Belgium, and South Tirol cannot simply be
ignored. We therefore know little about segmental
reduction behaviour in these regions and how it
correlates to the codified or accepted reduction forms
of the (northern) standard. We also know little about
regional variation in temporal organization strategies,
particularly the intrinsic tempo of spoken German,
which is essential to temporal and segmental
reduction processes. Finally, the entanglement of
these factors when speech tempo is changed needs to
be investigated.
2. THE MISSING LINK
2.1. Prosodic variation and space
The disadvantages of most of the studies considering
speech tempo and space [17 for an overview] are
twofold. Firstly, measurements of speech tempo are
mostly set globally, meaning as average values by
utterance. These measurements only find whether
there are tempo differences, but not where these
differences come from. Global measurements like
articulation or speaking rates using words, syllables,
or phones per time frame and including or excluding
pauses have one thing in common: they do not reflect
the details of varying temporal organisation
strategies. For two hypothetical realisations of the
same sentence from speakers of different regions we
could obtain the same articulation rate value, but
diverging temporal strategies. These could be, for
instance, different final lengthening, varying intrinsic
vowel or consonant durations, assimilations,
deletions, or combinations of these factors.
Therefore, we investigate both global statistical
differences and the particular reasons for these
differences.
Second, the conceptualisation of language space is
often limited to the categories of traditional dialect
classifications or even to countries or particular
states, hindering insight into spatial continua.
Moreover, traditional classifications of German
dialect areas [2, 16, 21] are based on data collected in
the late 19th century and are mainly based on
phonological and morphological features. For this
time span of about 140 years, one could likely assume
some areal changes. What relationship do these
traditional dialect areas have to prosodic variation in
standard-intended German? It cannot be taken for
granted that current variation in tempo-related
features of standard-intended German necessarily
coincide spatially with dated findings of phonological
and morphological variation. Therefore, we use maps
which are suitable for illustrating spatial continua and
can be compared to the traditional dialect areas.
2.2. Combining the factors
Dressler [5] mentions that variants caused by fast
speaking are often difficult to distinguish from
seemingly identical variants of a particular regional
dialect. He shows that Breton dialects exhibit
regionally varying tendencies of vowel centralisation
in fast speech (‘allegro’). These findings on the
relationship between area and reduction inspired us to
identify regional variation in temporal pronunciation
features in German-speaking Europe [10]. On the
basis of 67 predominantly Federal German locations
we can show (1) that there is regional variation for
both articulation rate (segments/sec) and degree of
reduction (measured as segment deletion ratio); and
(2) that although these factors are closely interwoven,
there is no simple correlation. The complex interplay
between speech tempo, phonetic reduction, and
regional variation needs to be studied in detail,
especially in order to understand the composition of
global measurements. This paper summarizes the
progress of our investigations within the SpuRD-
project.
3. SPURD-PROJECT
3.1. Aims of the project
The general aim of this project is to disentangle the
interwoven relationships between speech tempo,
reduction, and regional space, considering the whole
German language area. Therefore, in a first step we
test spatial variation for several measures of speech
tempo and reduction indices on a macroscopic level.
Since we can assume that this macroscopic variation
accumulates from microscopic variation phenomena,
we also study the regionally varying details of
temporal organization as well as reduction
preferences. Furthermore, we investigate how
reading faster affects these spatial structures.
3.2. Material
We use reading material from the German Today
corpus [4] which contains two separate readings of
the German version of The North Wind and the Sun,
which were read in a normal and a fast tempo. In total,
1,494 recordings are available from a younger (17–19
years, high school graduates) and an older generation
(50–60, adult education classes), males as well as
females, who originate from 195 locations in Austria,
Belgium, Germany, Liechtenstein, Luxembourg,
South Tirol, and Switzerland. However, the location
density differs for the age groups, 165 locations are
available for the younger generation and only 79
locations for the older generation. The subjects had
no professional speaker training. The recordings were
made in a formal interview situation during class in a
quiet classroom. The preliminary results presented
here are based on 274 alreadly transcribed recordings
of the younger males from 119 locations, which have
been presegmented automatically at the sound level
with WebMAUS [12] and manually corrected in Praat
[3].
3.3. Methods: Measurement and mapping
Apart from the well-known disadvantages of using
scripted and formal linguistic utterances, reading
material provides the clear advantage of comparing
variables one-to-one in the same environment.
Geospatial comparability can only be achieved
through this. The individual transcripts are compared
with the canonical transcript that corresponds to the
standard pronunciation. Thus it is possible to gain
insights into spatial variation in the same linguistic
elements, recognizing regional reduction strategies
and not only average differences for a whole text or
utterance. The accurate manual segmentation allows
for measuring the sound segments in duration, pitch,
intensity, and spectral qualities, as well as for global
measurements on the text level for a general
comparison.
Duration values for the sounds were automatically
extracted with a Praat script. Segments that were
labelled as slips of the tongue, as well as pauses that
were not occlusions of plosives, were excluded. The
maps presented in Figure 1 illustrate the spatial
distributions of the following measurements: duration
of articulation (ArtDur in sec) measures the time,
minus pauses and slips of the tongue, that the
speakers needed to read the text; articulation rate
(ARsyll in σ/sec) measures the number of syllables
realized per ArtDur; the reduction index segment
deletion ratio (SDR in %) indicates the percentage of
canonical sounds deleted.
We use ArcGIS [7] for mapping. The measured
and averaged values for every location are transmitted
to voronoi polygons to get a surface impression of the
spatial distributions. A local smoothing method is
further applied to flatten individual variation in the
data. Hence, each polygon contains not only the
values for the speakers of the location it represents,
but the median value of itself and its direct
neighbours. Finally, the data is classified into 10
classes with the natural breaks or Jenks algorithm
[16], which is less arbitrary with class limits,
reflecting instead variation within the data itself by
minimizing the intra-class variation and maximizing
the inter-class variation; this method helps to detect
variance due to geographical space.
4. ANALYSIS
4.1. Macroscopic patterns
Table 1 gives an overview of the macroscopic
measurements for the total data and subdivided for
normal and fast tempo. As expected, the ArtDur
decreases from normal to fast reading, while the
values for SDR and ARsyll increase. In every case the
ANOVAs show a highly significant influence of
tempo. More interesting is the fact that the respective
standard deviations increase with fast reading, indi-
cating greater variation, which is a reinforcement of
regional patterns for fast speech.
Table 1: ArtDur, SDR, and ARsyll, Mean and SD,
ANOVA by tempo.
Figure 1 shows the spatial distributions for the glo-
bal measurements of ArtDur (in sec, excl. pauses), the
SDR (in %) and the ARsyll (in σ/sec, excl. pauses) for
the normal and fast reading conditions. The
classification of the values using the natural-breaks
method makes it possible, despite significantly
changed values between normal and fast reading, to
map the variation inherent in the data. Figure 1 shows
that the spatial structures for the measurement are
highly congruent between tempos and sharper in
shape when read fast, which partly explains the higher
SD for the fast condition. This similarity of spatial
structures strongly suggests that there is regional
intrinsic variation with regard to temporal
organization and the degree of reduction. Figure 1
also shows that the continua of value distributions do
not correspond well with the black-marked borders
between the traditional German regional dialects.
This, in turn, suggests that prosodic variation in
standard-intended pronunciation probably also
shapes independent spatial structures that cannot be
predicted from the traditional dialect areas. Table 2
lists the results of the ANOVAs by dialect area. The
locations of the study area were assigned to the 6
large regional dialects (Low, Middle, and Upper
German, each divided into an Eastern and Western
part) according to traditional classifications [21]. For
SDR, both ANOVAs yield highly significant results,
while for ARsyll and ArtDur only those for normal
tempo condition are significant. But considering the
maps, the problem of this approach reveals itself. The
dialect classification and the distributions of the
mapped measurements rarely coincide. Rather, we
find large variation within the dialect groups,
especially for ARsyll and ArtDur. For SDR, the areas
coincide better, i.e. with less internal variation, which
in this case explains the high significance of the
ANOVAs.
Figure 1: Overview: ArtDur, SDR, and ARsyll for
the whole German language area in central Europe
(on the left: normal tempo, on the right: fast tempo).
The spatial patterns for ArtDur and ARsyll match
very well to each other, reflected in a strong negative
correlation of r=-0.88 (Table 3). Areas in which
speakers need more articulation time to read the text
are simultaneously areas with low syllable rates, and
vice-versa. However, such a correlation does not exist
ArtDur
(sec)
SDR
(%)
ARsyll
(σ/sec)
Mean SD Mean SD Mean SD
total 28.8 3.7 11.5 4 6.4 0.8
normal
31.5 2.2 9.9 3.2 5.8 0.4
fast 26.0 2.6 13.2 4.2 6.9 0.7
ANOVA
by
tempo
F(1,273)
=351.87,
p<.0001*
F(1,273)
=53.26,
p<.0001*
F(1,273)
=262.85,
p<.0001*
between the duration or tempo measures and the
reduction index. Here, the correlations are regionally
variable. For example, East Middle German shows
high ArtDur (and low ARsyll), and according to the
SDR values, low segmental reduction. East Upper
German in the South, on the other hand, shows
similarly low reduction values with on average
significantly lower ArtDur (and higher ARsyll).
Table 2: One Way ANOVAs by dialect region for
ArtDur, SDR, and ARsyll.
In addition, it is noticeable from Table 3 that the
correlation strength decreases in all cases with higher
speech rate, suggesting that temporal organization
and segmental reduction are intertwined but
nevertheless independent prosodic systems with
independent spatial distributions.
Table 3: Pearson product-moment correlations.
It is precisely this ‘dependent independence’ that
makes this object of research so complex. Moreover,
it is important to be aware that regions with
macroscopically identical or similar values for
articulation duration may have very different
microstructures. The language material used here
may not by any means capture all possible variation,
but due to the direct comparability of the utterances it
is appropriate to show reliable regional tendencies. In
the following section, some of the microscopic
variations are sketched, which in their sum lead to the
macro-variation illustrated above.
4.2. Microscopic patterns: Exploring the reasons
In addition to the above analyses of global measures,
different segmental, temporal, and spectral aspects
are now analysed, allowing us to find partial
explanations for geographical structures and detect
interactions of different factors. So far in the SpuRD-
project we have analysed the realisation of schwa
between nasals [9]. The syncope of schwa is
dependent on speech tempo, but there is also a
geographical impact. While in the north schwa has
disappeared in both tempi, it is preserved in the
Bavarian area. In the transition zone, the syncope of
schwa is more dependent on tempo, which means that
schwa in these regions is more likely to become part
of the reduction mass when the tempo is increased,
while the Bavarian area remains largely insensitive to
this form of reduction. Furthermore, we have studied
the temporal realisation of the fortis-lenis contrast [8].
We find a geographical distribution of areas with both
large and small temporal contrasts.
In both analyses, the spatial distribution is
partially coherent with dialectal structures. Further
analyses are planned, e.g. for the intrinsic duration of
sounds and the realisation of the opposition between
short and long vowels. We will try to identify those
elements by factor analysis that can explain the main
part of macroscopic variation. We also look at tempo-
induced changes in sound quality. Elsewhere we
show that vowel space size is geographically
distributed distinctly from traditional dialect
boundaries – and that an increase in tempo results in
a change of the vowel space which is dependent on
the size of the vowel space in normal tempo, insofar
as larger vowel spaces are reduced, while originally
small spaces are enlarged [19].
5. CONCLUSION
Overall, this paper has shown that reading tempo,
articulation rate, and segment reduction are clearly
correlated, but that their linkage is complex,
especially due to regional variability. Macroscopic
and microscopic analyses must go hand-in-hand, as
these relationships can only be disentangled with both
perspectives. Therefore, in addition to investigating
phonetics, the sociolinguistic dimension of this
phenomenon must also be considered.
Furthermore, new methods, especially from the
geosciences (such as hot spot analyses), should be
tested and adapted for our purposes in order to be less
dependent on both traditional and modern dialect
classifications. These classifications are unsuitable
for exploring the spatial structures of prosodic
variation, as they are based on other linguistic levels
and provide insufficient linguistic space for statistical
analysis.
6. REFERENCES
[1] Ammon, U., Bickel, H., Lenz, A. N. 2016.
Variantenwörterbuch des Deutschen. Die
Standardsprache in Österreich, der Schweiz und
Deutschland sowie Liechtenstein, Luxemburg,
ANOVA
ArtDur (normal) F(1,138)=3.28, p=.008*
ArtDur (fast) F(1,134)=1.72, p=.1351
ARsyll (normal) F(1,138)=2.64, p=.0261*
ARsyll (fast) F(1,134)=1.04, p=.3996
SDR (normal) F(1,138)=10.01, p<.0001*
SDR (fast) F(1,134)=6.86, p<.0001*
Pearson product-moment correlations
ArtDur : ARsyll (normal) r=-0.88, p<.0001*
ArtDur : SDR (normal) r=-0.38, p<.0001*
ARsyll : SDR (normal) r=0.23, p=.0068*
ArtDur : ARsyll (fast) r=-0.70, p<.0001*
ArtDur : SDR (fast) r=-0.34, p<.0001*
ARsyll : SDR (fast) r=0.15, p=.0131*
Ostbelgien und Südtirol sowie Rumänien,
Namibia und Mennonitensiedlungen. 2nd ed.
Berlin: de Gruyter.
[2] Behaghel, O. 1891 Geschichte der deutschen
Sprache. In: Hermann Paul (ed.), Grundriß der
germanischen Philologie. vol. l, Straßburg:
Trübner, 526–633.
[3] Boersma, P. and Weenink, D. 2017. Praat: doing
phonetics by computer. Version 6.0.27.
<http://www.praat.org/>
[4] Brinckmann, C., et al. 2008. German Today: an
areally extensive corpus of spoken Standard
German. Proc. LREC Marrakech.
[5] Dressler, W. U. 1975. Methodisches zu Allegro-
Regeln. In: Dressler, W. U., Mareš, F. V. (eds.),
Phonologica 1972, Salzburg: Wilhelm Fink,
219–235.
[6] Elspaß, S./Dürscheid, C. 2017. Areale Variation
in den Gebrauchsstandards des Deutschen. In:
Konopka, M., Wöllstein, A. (eds.):
Grammatische Variation. Empirische Zugänge
und theoretische Modellierung. Berlin/Boston:
de Gruyter, 85–104
[7] ESRI 2015. ArcGIS Desktop: Release 10.3.
Redlands, CA: Environmental Systems Research
Institute.
[8] Hahn, M. in print. Zwischen Prozess und
Produkt. Zur Lenisierung velarer Plosive im
Deutschen. In: Hahn, M., et al. (eds.), Dynamik
in den deutschen Regionalsprachen. Hildesheim:
Olms.
[9] Hahn, M., Siebenhaar, B. in print. Schwa un-
breakable Reduktion von Schwa im
Gebrauchsstandard und die Sonderposition des
ostoberdeutschen Sprachraums. In: Kürschner,
S., et al. (eds.), Methodik moderner Dialekt-
forschung. Hildesheim: Olms.
[10] Hahn, M., Siebenhaar, B. 2016. Sprechtempo
und Reduktion im Deutschen (SpuRD). In:
Jokisch, O. (ed.), Elektronische Sprachsignal-
verarbeitung 2016. Dresden: TUDpress, 198–
205.
[11] Kehrein, R. 2009. Dialektalität von Vorleseaus-
sprache im diatopischen Vergleich – Hörerurteil
und phonetische Messung. Zeitschrift für
Dialektologie und Linguistik 76 (1). 14–54.
[12] Kisler, T., et al. 2017. Multilingual processing of
speech via web services. Computer Speech &
Language 45. 326–347.
[13] Kleiner, S. 2011ff. Atlas zur Aussprache des
deutschen Gebrauchsstandards (AADG).
http://prowiki.ids-
mannheim.de/bin/view/AADG [Nov. 29, 2018].
[14] Kohler, K. J. 1995. Einführung in die Phonetik
des Deutschen, 2nd edn. Berlin: Schmidt.
[15] Krech, E.-M., et al. 2009. Deutsches Aussprache-
wörterbuch. Berlin, New York: de Gruyter.
[16] Lameli, A. 2013 Strukturen im Sprachraum.
Analysen zur arealtypologischen Komplexität
der Dialekte in Deutschland. Berlin, Boston: de
Gruyter.
[17] Leemann, A. 2016. Analyzing geospatial varia-
tion in articulation rate using crowdsourced
speech data. Journal of Linguistic Geography 4
(02). 76–96.
[18] Lindblom, B. 1963. Spectrographic study of
vowel reduction. JASA 35 (11), 1773–1781.
[19] Siebenhaar, B., Hahn, M. 2019. Vowel Space,
Speech Rate and Language Space. In Calhoun,
S., Escudero, P., Tabain, M., & Warren, P. (eds.),
Proceedings of the 19th International Congress
of Phonetic Sciences. Melbourne, Australia.
[20] Simmler, F. 1983. Konsonantenschwächung in
den deutschen Dialekten. In: Besch, W., et al.
(eds.), Dialektologie: Ein Handbuch zur deut-
schen und allgemeinen Dialektforschung, vol. 2,
Berlin, New York: de Gruyter, 1121–1129.
[21] Wiesinger, P. 1983. Die Einteilung der deutschen
Dialekte. In: Besch, W., et al. (eds.), Dialekto-
logie: Ein Handbuch zur deutschen und allge-
meinen Dialektforschung, vol. 2, Berlin, New
York: de Gruyter, 807–900.
[22] Zimmerer, F. 2009. Reduction in natural speech.
Dissertation, University of Frankfurt.
... With a geographically balanced dataset of German read speech in two speech rates [1,2,3] we will focus on the following questions: ...
... This difference would then have nothing to do with the opposition of dialect and standard. One could also argue that this distribution has more to do with the regional distribution of speech rate or segment reduction than with dialect areas [2,3]. But even these relationships cannot be established unambiguously. ...
Conference Paper
Full-text available
Acceleration of speech rate is often said to be correlated with a reduction of the vowel space. However, a monocausal explanation of the vowel space reduction by speech rate is surely too simplistic. With our regionally balanced database of a German text read in two reading tempi we present geolinguistic maps of • the different sizes of the vowel space and • different changes of the sizes of the vowel space when comparing normal and accelerated speech rates. These maps for the normal reading tempo show regional patterns of vowel space sizes for the long and short vowel system. Accelerating speech rate affects the vowel space size in regionally specific patterns. In addition, increasing reading tempo shows a surprising general effect: initially large vocal spaces are reduced while initially small vocal spaces are enlarged. Interestingly, vowel space size and change due to accelerating reading tempo does only limitedly reflect traditional dialect regions.
Conference Paper
Full-text available
Acceleration of speech rate is often said to be correlated with a reduction of the vowel space. However, a monocausal explanation of the vowel space reduction by speech rate is surely too simplistic. With our regionally balanced database of a German text read in two reading tempi we present geolinguistic maps of • the different sizes of the vowel space and • different changes of the sizes of the vowel space when comparing normal and accelerated speech rates. These maps for the normal reading tempo show regional patterns of vowel space sizes for the long and short vowel system. Accelerating speech rate affects the vowel space size in regionally specific patterns. In addition, increasing reading tempo shows a surprising general effect: initially large vocal spaces are reduced while initially small vocal spaces are enlarged. Interestingly, vowel space size and change due to accelerating reading tempo does only limitedly reflect traditional dialect regions.
Chapter
Full-text available
The questions of how speaking rate and phonetic reduction are correlated as well as how they interact in linguistic space are still unanswered up to this date. The project speaking rate and phonetic reduction in German (Sprechtempo und Reduktion im Deutschen SpuRD) aims to answer these questions by using reading material from the Deutsch heute corpus of the Institute for German Language (Institut für Deutsche Sprache IDS). With these data HAHN/SIEBENHAAR (2016) detected a gradual slope from south to north of the German speaking area for articulation rate (AR). The detailed analysis shows that this slope in AR has different substructures due to regional differences in the interplay of average articulation duration (AD) and segment deletion. Following these findings, the present article focuses on regional patterns of segmental reduction by analysing the realisation of the schwa in final syllables between nasals. For the presented spatial patterns of schwa reduction the (East) Upper German Area stands out particularly as exceptionally resistant against schwa reduction in several ways.
Chapter
Full-text available
Die Untersuchung bietet den ersten systematischen Zugang zu der Frage, inwiefern Sprechtempo und phonetische Reduktion im deutschen " Gebrauchsstan-dard " regionaler Variation unterliegen. Dazu werden in zwei intendierten Tempi ein-gelesene Aufnahmen des " Nordwind und Sonne "-Textes ausgewertet, die es ermög-lichen, einen direkten Vergleich zwischen den bisher 67 über den deutschsprachigen Raum verteilten Ortspunkten zu ziehen. Die vorläufigen Ergebnisse zeigen, dass so-wohl Sprechtempo (gemessen als phonetische Artikulationsrate in Lauten/s) als auch Reduktion (gemessen als prozentualer Segmentelisionsgrad) z. T. deutliche Raum-muster ausprägen. Für die Artikulationsrate kann ein tendenzielles Süd-Nord-Gefälle festgehalten werden, während für den Segmentelisionsgrad ein deutlicher Kontrast zwischen nördlichen und westlichen Regionen auf der einen und östlichen und süd-östlichen Regionen auf der anderen Seite besteht. Im Weiteren kann festgehalten werden, dass ähnliche Werte für die Artikulationsrate regional sehr unterschiedlich zustande kommen und dadurch ein komplexes Zusammenspiel mehrerer Faktoren freigelegt wird. Um diese ersten Befunde zu präzisieren und zu differenzieren, müs-sen in weiterführenden Untersuchungen sowohl die konkreten Ausprägungen regio-nalspezifischer Reduktionsmuster als auch die temporale Feinstruktur der fokussier-ten Areale genauer betrachtet werden.
Thesis
Full-text available
Article
Full-text available
How "High German" or "dialectal" do linguistically naive hearers judge comparable speech samples from a defined group of speakers to be? This article describes a study in which the dialectality of speech samples from the sixteen most important dialect formations (according to Wiesinger 1983) reading aloud was assessed in a total of eight auditory tests. The assessments were performed in each of the six major German dialect regions in order to test for possible influences due to the hearers' origins. Statistical analyses failed to detect any such influences. A statistically highly significant correlation was found between the perceived level of dialectality and dialectality measurements based on a quantification of the phonetic distance of the speech samples from the pronunciation norm. A diatopic comparison of the origins of the speakers revealed a simultaneous decline from south to north and from east to west in terms of perceived and measured dialectality.
Book
Die Strukturierung des Sprachraums ist ein zentraler Gegenstand der Variationsforschung, dessen Komplexität bis heute nicht vollständig erfasst ist. Ausgangspunkt des Buches ist die Erstellung eines datenbasierten regionalsprachlichen Modells der historischen Dialekte in Deutschland, das aus verschiedenen Perspektiven ausgewertet und erweitert wird. Auf diese Weise gelingt eine neue Sicht auf die raumstrukturellen Bedingungen der Dialekte v. a. im 19. Jahrhundert. Eine Besonderheit stellt der Einsatz von Methoden aus Bioinformatik und Geostatistik dar, die in der Regionalsprachenforschung weithin unberücksichtigt geblieben sind. Im Ergebnis trägt das Vorgehen zum Verständnis zahlreicher sprachgeographischer und sprachhistorischer Phänomene bei, etwa mit Blick auf den Ausgleich der Dialekte oder die Genese der deutschen Schriftsprache. Zudem werden neue Kriterien zur intersubjektiven Definition der Dialektlandschaft geliefert, die in ihrer Verknüpfung mit außersprachlichen Daten einen Brückenschlag zu kulturwissenschaftlichen Themenfeldern leisten. Die Studie begründet darauf aufbauend neue Forschungsansätze zur Erklärung räumlicher Handlungsmuster in Vergangenheit und Gegenwart, seien sie sprachlicher oder außersprachlicher Art.
Article
Most recent studies on the geographical distribution of acoustic features analyze comparatively few speakers and localities, both of which may be unrepresentative of the diversity found in larger or more spatially fragmented populations. In the present study we introduce a new paradigm that enables the crowdsourcing of acoustic features through smartphone devices. We used Dialäkt Äpp , a free iOS app that allows users to record themselves, to crowdsource audio data. Nearly 3,000 speakers from 452 localities in German-speaking Switzerland provided recordings; we measured articulation rates for these speakers using a metric based on duration intervals between consecutive vowel onsets. Results revealed distinct regional differences in articulation rate between major dialect regions and individual localities. The specification of 452 localities enabled analyses at an unprecedented spatial resolution. Results further revealed a robust effect of gender, with women articulating significantly more slowly than men. Both the geographical patterns and the effect of gender found in this study corroborate similar findings on Swiss German previously reported in a very limited set of localities, thus verifying the validity of the crowdsourcing framework. Given the application of this new framework, a large bulk of the discussion is devoted to discussing methodological caveats.
Article
A new software paradigm `Software as a Service' based on web services is proposed for multilingual linguistic tools and exemplified with the BAS CLARIN web services. Instead of traditional tool development and distribution the tool functionality is implemented on a highly available server that users or applications access via HTTP requests. As examples we describe in detail five multilingual web services for speech science operational since 2012 and discuss the benefits and drawbacks of the new paradigm as well as our experiences with user acceptance and implementation problems. The services include automatic segmentation of speech, grapheme-to-phoneme conversion, syllabification, speech synthesis, and optimal symbol sequence alignment.
Article
Measurements of formant frequencies and duration are reported for 8 Swedish vowels uttered by a male talker in three consonantal environments under varying timing conditions. An exponential function is used to describe the extent to which formant frequencies in the vowels reach their target values as a function of vowel-segment duration. A target is specified by the asymptotic values of the first two formant frequencies of the vowel and is independent of consonantal context and duration. It is thus an invariant attribute of the vowel. The results suggest an interpretation in terms of a simple dynamic model of vowel articulation.