BieColl - Bielefeld eCollections

Buschmeier, Hendrik

Small but important – Building metacommunicative interaction mechanisms
into conversational agents
Hendrik Buschmeier
Bielefeld Universität
Artificial conversational agents, such as voice assistants and conversational interfaces, are
now widely available.  They are primarily used casually in the home for convenience and
entertainment  purposes,  but more impactful  applications,  e.g.,  as assistive technologies  in
home-care settings, are becoming more relevant. It is thus important that they are easy to use
for all types of users. Despite the significant advances in automatic speech recognition and
natural  language  processing  technologies  over  the  past  decade,  the  actual  conversational
capabilities of these assistants are still limited. One reason is that much of the interactional
infrastructure that humans rely upon to smoothly make progress in conversation – surfacing
in seemingly ‘small’ phenomena such as feedback, turn-taking signals, or repair – are often
overlooked. In this talk,  I will  focus on computational models of the metacommunicative
mechanism  of  linguistic  and  multimodal  feedback  and  their  effects  on  interactions  with
conversational agents and will make the argument that research in conversational technology,
but  also  the  disciplines  working  on  human  speech,  should  pay  more  attention  to
metacommunicative and interactional mechanisms and phenomena.
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-990


English

Small but important – Building metacommunicative interaction mechanisms into conversational agents

Lorenzen, Janne

Roessig, Simo

Baumann, Stefan

Individual variability in the prosodic encoding of information status in German 
 
Janne Lorenzen, Simon Roessig, Stefan Baumann 
IfL Phonetik, University of Cologne, Germany 
 
Information status in German is encoded via prosodic prominence, such that new referents 
are realized prosodically most prominently, given referents least prominently and accessible 
referents lie in between these extremes [1]. Prosodic prominence is multifaceted, encompassing 
a variety of cues related to timing, spectral properties and the F0 contour. It would seem 
reasonable to assume that not every speaker makes use of every single cue in order to encode 
a given (pragmatic) contrast. Previous studies have found, for example, considerable inter-
individual variability in focus type marking [2]. 
In order to investigate different strategies in the prosodic encoding of information status, we 
collected data from 32 participants in an interactive reading task via Zoom. Participants were 
presented with eight stories that they read aloud for a partner, who then had to sort 
corresponding picture cards into the correct order. Stories consisted of four sentences, the third 
of which was the target sentence. It contained two target words, an indirect object followed by 
a direct object, which were either accessible through the preceding context or new (e.g., Unter 
anderem hat sie dem Maler [accessible] eine Waage [new] verkauft. ‘Among other things, she 
sold the painter a scale.’). 
The data were segmented via WebMAUS [3,4] and prosodically annotated following the 
DIMA guidelines [5]. Specifically, two levels of phrase boundaries (strong vs. weak) were 
labelled. For the purposes of this paper, these two levels were collapsed to create a variable 
describing either the presence or absence of a boundary after a target word. The tone 
annotations were further translated into GToBI [6] accent types, which were then assigned a 
prominence score based on perceptual judgements of different accent types collected by [7]. In 
addition, word and (accented) syllable duration of the target words were measured in 
milliseconds. Cases where the audio was distorted in the target word were discarded (n = 6). 
Target words immediately preceding a phrase boundary were also excluded from duration 
measures to avoid effects of final lengthening.  
Figure 1 shows the differences in the normalized averaged parameter values between new 
and accessible target words for every speaker. Values above the horizontal line thus indicate 
that a speaker uses the parameter in a way to make new target words more prominent than 
accessible ones. Each vertical line represents one speaker.  
Overall, accent type appears to be the most robust cue as the majority of speakers (n = 24) 
realize new referents with more prominent accent types than accessible ones. In contrast, less 
than half of the speakers (n = 15) use phrase boundaries as a cue to make the preceding new 
referent more prominent. Nineteen speakers produce longer referents when they are new, and 
17 speakers lengthen the accented syllables of new in comparison with accessible referents. 
Grouped by type of cues employed, the largest group of speakers (n = 10, upper left panel) 
marks prominence exclusively via accent type. On the other hand, the speakers that do not 
modulate accent type (n = 8, upper right panel) preserve the accessible/new contrast via 
durational cues and/or phrase boundaries. Only six speakers encode information status 
maximally redundantly via all four parameters considered here. 
To conclude, while prosodic prominence can be encoded by many parameters, only few 
speakers exploit the full range of these cues, which is in line with the finding by [2]. Speakers 
tend to reduce this redundancy by following individual strategies in the encoding of prosodic 
prominence. A perception study will need to verify how successful these strategies are on the 
side of the listener.  
 
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-995
 
Figure 1. Normalized differences between new and accessible target words averaged within speakers 
  
 
[1] Baumann, S., & Riester, A. Coreference, lexical givenness and prosody in German. In 
Hartmann, J., Radó, J., & Winkler, S. (Eds.), [Special Issue “Information Structure 
Triggers”], Lingua, 136, 16–37, 2013. 
[2] Cangemi, F., Krüger, M., & Grice, M. Listener-specific perception of speaker-specific 
productions in intonation. In Fuchs, S., Pape, D., Petrone, C., & Perrier, P. (Eds.), Individual 
differences in speech production and perception. Frankfurt: Peter Lang, 123–145, 2015. 
[3] Schiel, F. Automatic phonetic transcription of non-prompted speech. In Proceedings of the 
14th International Congress of Phonetic Sciences, San Francisco, USA, 607–610, 1999. 
[4] Kisler, T., Reichel, U. D., & Schiel, F. Multilingual processing of speech via web services. 
Computer Speech & Language, 45, 326–347, 2017. 
[5] Kügler, F., Baumann, S., & Röhr, C. T. Deutsche Intonation, Modellierung und Annotation 
(DIMA) – Richtlinien zur prosodischen Annotation des Deutschen. In Schwarze, C., & 
Grawunder, S. (Eds.), Transkription und Annotation gesprochener Sprache und 
multimodaler Interaktion. Tübingen: Narr, 23–54, 2022. 
[6] Grice, M., Baumann, S., & Benzmüller, R. German intonation in Autosegmental-Metrical 
Phonology. In Jun, S.-A., (Ed.), Prosodic Typology: The phonology of intonation and 
phrasing. Oxford: Oxford University Press, 55–83, 2005. 
[7] Baumann, S., & Röhr, C. T. The perceptual prominence of pitch accent types in German. 
Proceedings of the 18th ICPhS, vol. 298, Glasgow: Univ. of Glasgow, 1–5, 2015.  
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-995


Individual variability in the prosodic encoding of information status in German

Youssef-Baronfeind, Cassandra

Differences in VOT in L1 Arabic and L3 German

Nebert, Uli

Mündlich gendern. Aber in welcher Form?  
Geschlechtsneutrale Sprache ist umstritten, hat sich aber im Schriftlichen mehr oder 
weniger etabliert (vgl. Beschluss des Rats für deutsche Rechtschreibung 2018). Es 
folgt der Grundannahme, dass Sprache die Art und Weise der Wahrnehmung 
beeinflusst (vgl. u. a. Braun et al. 2007) und man mit „gerechter Sprache“ zeigen kann, 
dass eine gerechtere Welt gewollt wird (vgl. Stefanowitsch 2018). Dies bezieht jedoch 
die mündliche Sprache mit ein, die den Großteil unseres Sprechhandeln ausmacht.  
Um festzustellen, welche Formen dabei welche Akzeptanz finden, fand eine 
Untersuchung statt. Dabei hörten unterschiedliche Probanden jeweils eine Rede in 
einer der unterschiedlichen Formen im Umgang mit dem Genus:  
1. Verwendung des „traditionellen“ generischen Maskulinums  
2. Verwendung des jeweils weiblichen und männlichen Genus mit entsprechender 
Movierung  
3. Verwendung der Endung „*innen“ (Glottisschlag nach Wortstamm mit 
zusätzlicher Betonung)   
4. Umgehung der Genusnennung, insbesondere durch die inhaltlich passende 
Substantivierung des Partizips  
Durch die Aufstellung anschließender Fragen kann auf die Reproduktionsfähigkeit bei 
den unterschiedlichen sprachlichen Formen geschlossen werden. Außerdem wurde 
die Akzeptanz für eine Standardisierung abgefragt sowie das Geschlechterverhältnis 
der Erwähnten. So können Rückschlüsse auf das Verständnis und das von den 
Probanden empfundene sprachliche Diskriminierungspotential innerhalb des 
Untersuchungsmaterials gezogen und die genannten unterschiedlichen Formen der 
Verwendung untereinander verglichen werden. Die aktuelle Untersuchung fand im Juni 
2022 statt, die Ergebnisse werden präsentiert.  
  
Literatur:  
AG Feministisch Sprachhandeln Was tun? Sprachhandeln – aber wie? W_Ortungen 
statt Tatenlosigkeit! (2. Aufl.) Berlin: AG Femisitisch Sprachhandeln, 2015. 
Braun, F., Oelkers, S., Rogalski, K., Bosak, J., Sczesny, S. Aus Gründen der 
Verständlichkeit ...“ Der Einfluss generisch maskuliner und alternativer 
Personenbezeichnungen auf die kognitive Verarbeitung von Texten. In: 
Psychologische Rundschau 58 (3), 183–189, Bern: Hogrefe, 2007. 
Stefanowitsch, A. Eine Frage der Moral: Warum wir politisch korrekte Sprache 
brauchen. Berlin: Duden, 2018. 
Rat für deutsche Rechtschreibung. Empfehlungen zur „geschlechtergerechten  
Schreibung“. Beschluss des Rats für deutsche Rechtschreibung vom 16. November 
2018. Mannheim: Institut für deutsche Sprache, am 1. 9. 2022: 
https://www.rechtschreibrat.com/DOX/rfdr_PM_2018-11-
16_Geschlechtergerechte_Schreibung.pdf .  
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1003


German

Mündlich gendern - Aber in welcher Form?

Choi, Yeongeun

 
Durational cues of Standard German Plosives: 
A comparison between Austrian, German, and Swiss speakers 
 
Yeongeun Choi 
Department of Computational Linguistics, University of Zurich 
 
The present study contrastively analyzes the acoustic cues of word-initial plosives in 
Standard German, as pronounced by Austrian, German, and Swiss speakers. German belongs 
to the so-called ‘aspirating’ languages [1] and distinguishes homorganic lenis and fortis 
plosives in word-initial position mainly along the short-to-long lag Voice Onset Time (VOT) 
spectrum [2, 3]. The role of closure duration (CD) and of fundamental frequency effects (CF0) 
of the vowels immediately following the plosives is not entirely clear yet [4, 5]. Whereas the 
standard plosives in the varieties spoken in Austria and in Germany have been analyzed 
separately [2, 3, 6, 7], no work has been done on Swiss Standard German so far; moreover, 
there are no previous studies which compare plosives in the three national varieties of Standard 
German. In this contribution we investigate the durational contrasts of German plosives 
realized in terms of VOT and CD. We hypothesize that German speakers will most clearly 
distinguish contrasting homorganic plosives by effectively using both VOT and CD (H1), 
whereas Austrian speakers will make less use of the two acoustic cues for plosive 
discrimination (H2). As regards Swiss speakers, we expect to find smaller VOT differences, 
whereas CD might play a major role considering their dialectal background [8] (H3). 
A reading task with a list of German interrogative sentences was conducted, where the target 
word containing a word-initial plosive is under focus (e.g., /t/ in Magst du lieber tanzen oder 
singen?), following the procedure used by [8]. Three groups of university students were 
recruited in Frankfurt, Vienna, and Zurich; ten native speakers from each group have been 
analyzed (mean age = 22, age range 18-29 years). All subjects are female, except for two male 
German speakers and one male Swiss speaker. The speakers were recorded producing a total 
of 48 interrogative sentences (3 places of articulation x 2 phonation types x 2 vowels x 4 
repetitions). The vowels preceded by the initial plosives of the target word were restricted to 
the low vowels <a> and <o> only. Overall, a total of 1,440 tokens (48 words x 30 subjects) 
were analyzed. The target acoustic features of the initial plosives were manually segmented 
and annotated in Praat [9]. The proportions of VOT and CD durations relative to the duration 
of the target word were calculated in order to control the speech rate of the collected recordings, 
as proposed in [3]. 
The results of VOT proportions show that both plosive type and place of articulation have 
a significant effect in each of the three groups. German speakers most effectively used VOT as 
an acoustic cue to distinguish lenis and fortis plosives (see Fig. 1). Regarding place of 
articulation, the more anteriorly the plosives were articulated, the smaller the VOT proportions 
were for all subject groups, except for the velar fortis plosives of German speakers. CD 
proportions also appear to be influenced by both plosive type and place of articulation, but 
differed in the three groups. In general, lenis stops had higher CD proportions than fortis stops 
in Austrian and German speakers, but the opposite occurred in Swiss speakers. The effect of 
place of articulation was shown to vary depending on the plosive type. For fortis stops, the CD 
proportions decreased from bilabial to velar stops (see Fig. 2). In lenis stops, however, the 
length of the articulatory tract seems to influence CD proportions only in German and Swiss 
speakers, but not in Austrian speakers. 
So far, the results confirm the three hypotheses as regards VOT, whereas for CD the 
situation is less clear; at least, Austrian and Swiss speakers seemed to make a slightly different 
use of CD. All in all, however, university students seem to pronounce the plosives of Standard 
German in a rather similar way. 
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1015
 
Figure 1. VOT proportions (percentage of 
lenis VOT in relation to fortis VOT) across the three 
German varieties 
Figure 2. CD proportions (percentage of lenis CD 
in relation to fortis CD) across the three German 
varieties 
    
 
 
References 
 
[1] Beckman, J., Jessen, M., & Ringen, C. Empirical evidence for laryngeal features: Aspirating 
vs. true voice languages. Journal of Linguistics, 49, 259–284, 2013. 
[2] Kleber, F. VOT or quantity: what matters more for the voicing contrast in German regional 
varieties? Results from apparent-time analyses. Journal of Phonetics, 71, 468-486, 2018. 
[3] Luef, E. M. Development of voice onset time in an ongoing phonetic differentiation in 
Austrian German plosives: Reversing a near-merger. Zeitschrift für Sprachwissenschaft, 
39(1), 79–101, 2020. 
[4] Kirby, J., Kleber, F., Siddins, J., & Harrington, J. Effects of prosodic prominence on 
obstruent-intrinsic F0 and VOT in German. In Minematsu, N., Kondo, M., Arai, T. & 
Hayashi, R. (Eds.), Proceedings of the 10th International Conference on Speech Prosody 
2020. Tokyo: ISCA, 210-214, 2020. 
[5] Zellers, M., & Schuppler, B. Microprosodic Variability in Plosives in German and Austrian 
German. In Meng, H., Xu, B. & Zheng, T. F. (Eds.), Proceedings of Interspeech 2020. ISCA. 
https://dx.doi.org/10.21437/Interspeech.2020-2353, 656-660, 2020. 
[6] Moosmüller, S., & Ringen, C. Voice and aspiration in Austrian German plosives. Folia 
Linguistica, 38(1-2), 43-62, 2004. 
[7] Hödl, P. Production and perception of voice onset time in Austrian German (Ph.D. 
dissertation). Graz: Karl-Franzens-Universität Graz, 2019. 
[8] Ladd, D. R., & Schmid, S. Obstruent voicing effects on F0, but without voicing: Phonetic 
correlates of Swiss German lenis, fortis, and aspirated stops. Journal of Phonetics, 71, 229-
248, 2018. 
[9] Boersma, P., & Weenink, D. Praat: doing phonetics by computer. Computer program, 
Version 6.1.52, retrieved 25 August 2021 from http://www.praat.org/. 
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1015


Durational cues of Standard German Plosives: A comparison between Austrian, German, and Swiss speakers

Nieder, Jessica

Tomaschek, Fabian

Maltese morphophonology – a merger of two worlds
Jessica Nieder & Fabian Tomaschek
Heinrich-Heine-Universität Düsseldorf, Eberhard Karls University of Tübingen
From a linguistic perspective, the Semitic language Maltese is situated between two worlds: Ro-
mance, especially Sicilian Italian, vs. Semitic, especially Maghrebi Arabic varieties (Comrie,
2007). The “hybrid” morphophonological system (Borg & Gatt, 2017; Mayer et al., 2013) is, for
example, visible in the Maltese plural formation: A variety of concatenative sound plurals, such
as omm-ommiijiet ‘mothers’, characterised by additonal phonological material at the end of the
word form contrasts with a variety of non-concatenative broken plurals, such as kelb-klieb ‘dogs’,
characterised by a stem alternation.
In this study we present a computational modeling approach of this hybrid morphophonolog-
ical system by using Naive Discriminative Learning (NDL), an error-driven learning algorithm
predicting so-called outcomes based on form-based cues (e.g. phones), to classify Maltese nouns
as belonging to the language families Semitic vs. Non-Semitic. NDL was trained on a data set of
2377 Tunisian nouns from Gugliotta and Dinarelli (2020), 2377 randomly selected Italian nouns
from the subtlex-it corpus1 and 2377 randomly selected German nouns from the CELEX corpus
(Baayen et al., 1996). The model was then tested on 6511 Maltese singular and sound and broken
plural nouns from Nieder et al. (2021). We used the same quasi-phonetic transcription based on
SAMPA for all three languages. 2-phones were used as phonological cues, the language families
semitic vs. non-semitic were used as outcome.
If Maltese is indeed a merger the two linguistic worlds Romance & Semitic and the knowl-
edge of native speakers of this merged system is based on error-driven learning, we expect NDL
to show high classification accuracies for a training based on Tunisian and Italian data and lower
accuracies for a training based on Tunisian and German data. Table 1 shows the accuracies for
the tested models. The evaluation on the training data (first row) showed an excellent overall
accuracy of >97%, indicating that NDL is able to capture the phonological differences between
Semitic (Tunisian) and Non-Semitic (Italian, German) languages and to accurately classify nouns
accordingly.
For the Maltese test data (second row), we found that the network predicts the origin of the
word forms with a high accuracy of 72% for a training based on Tunisian and Italian. For a training
on Tunisian and German, the accuracy drops to 59% and thus confirms our initial hypothesis.
Interestingly, when having a closer look at the Maltese test nouns that were to be classified,
we discovered that within the group of nouns with a Semitic origin, NDL showed overall high
classification accuracies with >80% across all number categories (singular, broken, sound). For
nouns with a Non-Semitic origin, NDL was unsure how to classify singulars and broken plural
nouns. This uncertainty is a direct reflection of the hybrid morphological system with nouns such
as denfil (sg.) - dniefel (broken pl.), etymologically related to Italian delfino ‘dolphin’, following
a non-concatenative plural pattern despite having a Non-Semitic origin.
Our results show that it is possible to classify Maltese nouns using NDL with a training based
on Tunisian and Italian nouns only. The model successfully applied the language-specific knowl-
edge gained during the training to Maltese nouns, thus confirming the hybrid status of the Maltese
morphophonological system.
1downloaded from http://crr.ugent.be/programs-data/subtitle-frequencies
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1018
Italian+Tunisian German + Tunisian
accuracy training 97.35 97.23
accuracy test 71.59 58.96
Table 1: Accuracies of NDL models trained on Italian + Tunisian data and trained on German +
Tunisian data.
References
Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1996). The celex lexical database (cd-rom).
Borg, C., & Gatt, A. (2017). Morphological analysis for the Maltese language: The challenges of a hybrid system.
Proceedings of the Third Arabic Natural Language Processing Workshop, 25–34. https://doi.org/10.18653/
v1/W17-1304
Comrie, B. (2007). Maltese and the World Atlas of Language Structures. Introducing Maltese Linguistics: Selected
Papers from the 1st International Conference on Maltese Linguistics (pp. 3–15). John Benjamins Publishing.
Gugliotta, E., & Dinarelli, M. (2020). TArC: Incrementally and Semi-Automatically Collecting a Tunisian Arabish
Corpus. LREC.
Mayer, T., Spagnol, M., & Schönhuber, F. (2013). Fixing the broken plural in Maltese. Perspectives on Maltese
Linguistics (pp. 129–158). De Gruyter.
Nieder, J., Chuang, Y.-Y., van de Vijver, R., & Baayen, R. (2021). Comprehension, Production and Processing of
Maltese Plurals in the Discriminative Lexicon. https://doi.org/https://doi.org/10.31234/osf.io/rkath
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1018


Maltese morphophonology – a merger of two worlds

Arkhipov, Alexandre

Stress patterns interaction in loanwords in Kamas 
 
Alexandre Arkhipov 
Universität Hamburg 
 
We will examine word prosody patterns in Russian borrowed nouns in Kamas compared to 
native items and earlier Turkic loans, focussing on vowel duration and tonal movement locus. 
Kamas is an extinct Samoyedic (< Uralic) language. It used to be spoken in Southern 
Siberia; the last speaker, Klavdiya Plotnikova, died in 1989. This study is based on the INEL 
Kamas Corpus [1] which contains transcripts of all known recordings of Plotnikova (ca. 14 
hours total; see [2] for details). Kamas had extensive contacts with Siberian Turkic languages, 
which probably lasted for a few centuries [3]. The next major contact language was Russian; 
in the 20th century, the remnants of the Kamas community shifted to Russian completely. 
Plotnikova’s Kamas is affected by language erosion as well as Russian influence on all levels, 
however Kamas phonetics is relatively intact in her speech. 
Kamas word prosody system before the language shift to Russian was described by Donner 
[4] as follows. The primary stress (expressed by an ‘expiratory accent’) would typically fall on 
the first syllable. The primary stress could shift to the second syllable if it was closed or had a 
long vowel. Under Turkic influence, the word-final syllable could acquire prominence, not 
only in Turkic loans but also in native lexemes. As Klumpp notes [3], “[t]he Turkic ultima 
accent, however, is a pitch accent, and we don’t know how much it co-occurred with the 
previous patterns”. Note that vowel length in Kamas is automatic in some contexts (e.g. before 
/r/: par- [paːr-] ‘return’) and distinctive in others (e.g. when originating from vowel contraction, 
as in kan- ‘go’ vs. kaːn ‘Khan’ (< Turkic qaɣan ‘Khan’). 
In Plotnikova’s recordings, we can observe the following patterns. First, there is no apparent 
first-syllable prominence tendency. In uninflected native Kamas words and older (Turkic) 
loans, it is the final syllable which typically shows greater vowel duration, unless a preceding 
syllable has a long vowel (e.g. nüke ‘woman’ vs. büːzʼe ‘man’). If there is a (phrase-level) F0 
movement, it is also usually anchored to the final syllable, whatever the vowel duration. 
Inflectional suffixes differ in their stress-related properties. The plural suffix -ʔi, for instance, 
acts as a stress attractor: it most often bears a very prominent tonal movement, and in many 
cases also shows vowel lengthening. (In what follows, we will mark this with acute, e.g. ine-ʔí 
‘horse-PL’). It can keep both properties when followed by a case marker, e.g. instrumental -zi 
or lative -nə/-də: ine-ʔí-zi ‘horse-PL-INS’. This is not the case with the other, restricted plural 
suffix -zaŋ; when followed by a case suffix, it is the latter which bears the tonal movement, cf.: 
tumo-ʔí-nə ‘[mouse-PL-LAT] to the mice’ vs. măja-zaŋ-də́ ‘[mountain-PL-LAT] to the 
mountains’. The root vowel length is generally unaffected by the inflectional suffixes. 
Now consider Russian loanwords. In uninflected wordforms, the original Russian stress is 
typically manifested in vowel lengthening, without marked tonal changes: kapuːsta ‘cabbage’ 
(Rus. kapústa). (Note that while vowel duration in Russian is considered a major cue to stress, 
it is also the stressed syllable which serves as anchor to phrasal tonal movements.) When 
Kamas inflectional suffixes are added to a Russian noun, they show the same stress-related 
properties as in Kamas words, while the vowel lengthening representing the Russian stress is 
by default intact. E.g.: alʼeːnʼə-ʔí ‘reindeer-PL’, sʼestraː-zaŋ-də́ ‘[sister-PL-POSS] her sisters’.  
However, there is evidence of interaction between the two systems. Apparently, the situation 
when the lengthening (marking the Russian stress) is adjacent to the Kamas stressed suffix is 
avoided. One option is to shift (Russian) stress to the left, as in groːza-ʔí ‘thunderstorm-PL’ 
(cf. Rus. grozá [ɡrɐˈza]), whereby the neutralized root vowel is restored. Another option is to 
omit lengthening altogether, as in sasna-ʔí ‘pine-PL’ (Rus. sosná [sɐsˈna]). Another factor 
which comes into play is the avoidance of the unstressed high central [ɨ], atypical in Kamas. 
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1025
 
[1] Gusev, V., Klooster, T., & Wagner-Nagy, B. 2019. INEL Kamas Corpus. Version 1.0. 
Publication date 2019-12-15. http://hdl.handle.net/11022/0000-0007-DA6E-9. Archived in 
Hamburger Zentrum für Sprachkorpora. In: Wagner-Nagy, B., Arkhipov, A., Ferger, A., 
Jettka, D., & Lehmberg, T. (eds.). The INEL corpora of indigenous Northern Eurasian 
languages. 
[2] Arkhipov A., Däbritz C. L., & Gusev V. User’s Guide to INEL Kamas Corpus. Working 
Papers in Corpus Linguistics and Digital Technologies: Analyses and Methodology 3. 
Szeged & Hamburg: University of Szeged, Universität Hamburg, 2020. 
[3] Klumpp, G. Kamas. In: Bakró-Nagy, M., Laakso, J., & Skribnik, E. The Oxford Guide to 
the Uralic Languages. Oxford: Oxford University Press. P. 817–843, 2022. 
[4] Joki, Aulis J. Kai Donners Kamassisches Wörterbuch nebst Sprachproben und Hauptzügen 
der Grammatik. Lexica Societatis Fenno-Ugricae VIII. Helsinki: Suomalais-Ugrilainen 
Seura, 1944. 
 
This publication has been produced in the context of the joint research funding of the 
German Federal Government and Federal States in the Academies’ Programme, with funding 
from the Federal Ministry of Education and Research and the Free and Hanseatic City of 
Hamburg. The Academies’ Programme is coordinated by the Union of the German Academies 
of Sciences and Humanities. 
P&P 18, 6-7 October 2022, Bielefeld
Licensed under CC BY 4.0
doi:10.11576/pundp2022-1025


Stress patterns interaction in loanwords in Kamas

Schubert, Dirk

Pohlan, Jörg

Lasch, Hülya

Hein, Carola

Höller, Lukas

Lofi, Christoph

Chu, Lixia

      
 
Median income relationships with Corona incidence in three waves of the COVID-19 pandemic in municipalities3
Incidence of Covid-19: Cases/Day per 100.000 Inhabitants in Municipalities (Summarized Time Series Data by 6 Months)2 
Daily confirmed new cases and deaths of Coronavirus (Covid-19) in Germany from January 2020 to September 20211
TIME
SPATIAL
INCOME
MUNICIPALITIES
        01.03.20 - 31.08.20        01.09.20 - 28.02.21         01.03.21 - 31.08.21                            
Incidence of Covid-19 
cases per 100 000 
inhabitants
6 Monthly
COVID-19
Pearson 
correlation
between 
Covid-19
Introduction
Goals
Methods
Socio-cultural factors of health
First results
Spatial, temporal and social analysis
Time Travel Project has identi-
fied more than hundred indi-
cators that can be explored in 
relation to corona-related data 
sets (the spread and develop-
ment of the Covid-19 disease 
over time and through space, 
the impact of vaccination rela-
ted to the built environment)
Time Travel - Crossing socio-cultural datasets and the Covid-19 pandemic in space over time 
Foundation:
Analysis of
rela�onships
Corona & Social
Experimenta�on 
of new methods
Iden�fica�on of the 
most affected ci�es 
& neighborhoods
Conclusion and next steps
Caution with Interpretation!
Emprical social & 
explorative research
Secondary analysis of 
quantitative data
  
Geodata visualization with 
the help of Geographic 
Information Systems (GIS)
Machine Learning models
To help provide new (planning) strategies for 
healthier urban environments, this project 
proposes to display diverse non-pre-deter-
mined spatial, socio-cultural as well as 
Corona-related datasets (Covid-19 cases, 
vaccinations) in space and time. Our research 
aims to overcome challenges of existing rese-
arch due to biases, and lack of detail that 
result from a varying and inconsistent (pre-) 
selection of static, singular, or limited so-
cio-cultural data. The main purpose of this 
project is to ascertain the relationships bet-
ween health factors and Covid-19-related 
features (outcomes) by visualizing and analy-
zing crossed datasets in space over time in a 
data-driven approach.
• Spatial analysis of corona-related data leads to 
better understanding of the role of the built environ-
ment on the spread of the corona virus and the 
impact of vaccination 
• Temporal analysis leads to better understanding of 
the dynamics and complexity of the evolution of the 
spread of the Covid-19 disease and vaccinations
• Social analysis leads to better understanding exis-
ting socio-cultural factors on the spread of Covid-19 3 Phases 
(6 Monthly) of 
the Corona desease 
from March 2020 
to the end of 
August 2021
• Explorings: Corona incidences - phases/waves - spread of Covid-19 – socio-cultural factors
• At the beginning phases of pandemic were found different spatial relationships between 
socio-cultural factors and the spread of Covid-19 comparing to later stages of the pandemic. 
• Graphics show spatial relationships between Corona incidences and income in different 
phases of pandemic as an example to the provided research for Germany and Netherlands.
SO
CIA
L
• There is not automatically a 
causality just because of an 
identified relationship/pattern 
between factors
• Research risk for discovering 
unmeaningful overlaps and chal-
lenging discovered results to 
verify and draw conclusions
• The selection of methods and 
scales provides diverse insights
Developing a Machine Learning methodology based on 
non-predetermined datasets can help reveal evidence of 
unknown relations between corona and socio-cultural fac-
tors, reduce health-related socio-cultural inequalities and 
promote healthier environments.
• Translating those new outcomes to arrive at planning 
and policy strategies for healthier cities
• Pilot study on spatial, temporal and social data for 
Hamburg Metropolitan Area (Germany) and the Province 
of South-Holland (The Netherlands)
• Identification of the most affected cities and regions on 
Covid-19 in relation to the socio-cultural factors
HafenCity University: Prof. Dr. Dirk Schubert, Prof. Dr. Jörg Pohlan, M.Sc. Hülya Lasch
Delft University of Technology: Prof. Dr. Ing. Carola Hein, M.Sc. Lukas Höller, Dr. Christoph Lofi, Dr. Lixia Chu
Team Contact:
HafenCity University: Prof. Dr. Dirk Schubert - Department: Housing & Neighborhood Development E-Mail: dirk.schubert@hcu-hamburg.de, Prof. Dr. Jörg Pohlan - Department: Urban 
planning und quantitative methods E-Mail: joerg.pohlan@hcu-hamburg.de
Delft University of Technology: Prof. Dr. Ing. Carola Hein - Professor of History of Architecture and Urban Planning - Department of Architecture
E-Mail: C.M.Hein@tudelft.nl, M.Sc. Lukas Höller - PhD Student Architecture and the Built Environment Department of Urbanism, Section of Spatial Planning & Strategy
E-Mail: l.holler@tudelft.nl, Dr. Christoph Lofi - Assistant Professor, Department: Web Information Systems group of the Faculty of Engineering, Mathematics and Computer Science 
(EEMCS/EWI) E-Mail: c.lofi@tudelft.nl, Dr. Lixia Chu - Postdoctoral Researcher in Big Earth Observation data cloud-computation, long-term EO time series, GIS and human-environment 
sustainable development Departments: Electrical Engineering, Mathematics and Computer Science -Software Technology- Web Information Systems E-Mail: L.Chu-1@tudelft.nl
First Author Contact:
M.Sc. Hülya Lasch 
HafenCity University Hamburg
Research assistant, by project Time Travel – Urban Planer & 
GIS Scientist / Urban planning and quantitative methods
Adress: Henning-Voscherau-Platz 1 - 20457 Hamburg
Room: 4102 Tel. +49 40 428 27 4011
huelya.lasch@hcu-hamburg.de
www.hcu-hamburg.de
Sources:
1 Own illustration according to Statista /RKI (Stand: 8. September 2021) 
https://de.statista.com/statistik/daten/studie/1100739/umfrage/entwicklung-der-taeglichen-fallzahl-des-coronavirus-in-deutschland/#:~:text=Am%2014.,auf%2
0mehr%20als%20197%20Millionen.
2 Own illustration according to RKI History/ COVID-19 Datahub ESRI 
https://npgeo-corona-npgeo-de.hub.arcgis.com/datasets/6d78eb3b86ad4466a8e264aa2e32a2e4_0/about
3 Own illustration according to Corona-Datenplattform Infas https://www.corona-datenplattform.de/dataset/?_tags_limit=0&tags=finanzen
Icon images: Flaticon.com and thenounproject.com
r: 0,37 r: -0,20 r: 0,04
Identification of the 
most affected cities 
and regions in relation 
to the socio-cultural 
factors and deriving a 
transferable framework  
1st wave 3th wave and start of 4th wave2nd wave
1st wave 2nd wave 3th wave
Daily new cases/deaths
Analysis of the 
relationships 
between the Corona 
and socio-cultural 
factors to identify 
their significance
Examination and 
implementation of 
new methods to 
draw evidence-based 
conclusions from 
large datasets
Phases
new cases  deaths   
1st wave 2nd wave 3th wave


Time Travel - Crossing socio-cultural datasets and the Covid-19 pandemic in space over time

Kohler, Martin

Stokmann, Antje

Rost, Amelie

Foto: Daniel Lerps 
Urbane Gewässer als flüssiger Freiraum für Alle
Martin Kohler, Antje Stokmann, Amelie Rost
Foto: Herm Holten
Foto: charmcity123

§ 9 Gemeingebrauch im Allgemeinen
(1) Jeder darf unter den Beschränkungen des § 23 WHG oberirdische Gewässer und 
die Küstengewässer zum Baden, Waschen ohne Verwendung wassergefährdender 
Stoffe, Schöpfen mit Handgefäßen oder Motorpumpen mit geringerer Leistung als 0,25 
Kilowatt, und zum Eissport, mit Ausnahme des Eissurfens und des Eissegelns, 
benutzen. 
§ 10 Schifffahrt
(1) Schiffbare Gewässer dürfen im Rahmen des Schifffahrtrechts einschließlich des 
Schifffahrtabgabenrechts von jedermann mit Wasserfahrzeugen befahren werden. 
Welche Gewässer schiffbar sind, bestimmt der Senat durch Rechtsverordnung.
(2) Nicht schiffbare Gewässer dürfen mit kleinen Wasserfahrzeugen ohne maschinellen 
Antrieb befahren werden. Der Senat kann durch Rechtsverordnung das Befahren mit 
kleinen maschinell angetriebenen Wasserfahrzeugen zulassen.
Hamburger Wassergesetz HaWG
§ 16 Benutzungsbedingungen und Auflagen
(1) Benutzungsbedingungen und Auflagen sind insbesondere zulässig, um
1. nachteilige Wirkungen zu verhüten und auszugleichen
a) für den Wasserhaushalt und die Trinkwasserversorgung,
b) für Natur und Landschaft, insbesondere die Pflanzen- und Tierwelt,
c) für die Sicherheit und Ordnung auf den Gewässern,
d) für die Gesundheit der Bevölkerung, das Wohnungs- und Siedlungswesen, das Stadtbild und den Hochwasserschutz,
e) für die gewerbliche Wirtschaft, die Land- und Forstwirtschaft, den Gartenbau, den Verkehr, den Bergbau und die 
Fischerei,
2. die sparsame Verwendung des Wassers sicherzustellen.
§ 27 c Fristen zur Erreichung der Bewirtschaftungsziele
(1) Ein guter ökologischer und chemischer Zustand der oberirdischen Gewässer im Sinne des §§ 25 a Absatz 1 Nummer 2 
WHG, ein gutes ökologisches Potential und ein guter chemischer Zustand der künstlichen und erheblich veränderten 
Gewässer im Sinne des § 25 b Absatz 1 Satz 1 Nummer 2 WHG sowie ein guter mengenmäßiger und chemischer Zustand 
des Grundwassers im Sinne des § 33 a Absatz 1 Nummer 4 WHG sind bis zum 22. Dezember 2015 zu erreichen.
Hamburger Wassergesetz HaWG
Unterschiedlich
Unsichtbar
Unzuständig
Können aber 
sinnvoll nur 
gesamtheitlich 
geplant werden.
Wasserräume sind…
Quelle: Gunnar Spreen, Universität Bremen
Foto: Amelie Rost
Gute Beispiele: Kulturfloss Schaluppe
https://mobilemachenschaften.de/schaluppe/
Quelle: IBA Hamburg
Gute Beispiele: Tidepark Elbe / Pilotprojekt Kreetsand
https://www.internationale-bauausstellung-hamburg.de/projekte/deichpark-
elbinsel/pilotprojekt-kreetsand/projekt/pilotprojet-kreetsand.html
Quelle: Martin Kohler
Gute Beispiele: Künstlicher Wasserlandschaftspark durch gezielte Verlandung mit 
ökologischen Funktionen eines tidebeeinflußten Auenwalds mitten in Hamburg
Weitere Informationen: martin.kohler@haw-hamburg.de
Foto: Hallo e.V.
Vielen Dank!
martin.kohler@haw-hamburg.de


Urbane Gewässer als flüssiger Freiraum für Alle

Mekel, Odile

Internationale Ansätze für StadtGesundheit

Not a member yet