The Future of Prosody Its about Time Dafydd Gibbon Bielefeld - - PowerPoint PPT Presentation

The Future of Prosody

It’s about Time

Dafydd Gibbon

Bielefeld University Jinan University Speech Prosody 9, Poznań, 13 June 2018

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

2

TIME

The Future of Prosody – relevant topics?

Ethics of research

• responsibility for the use of our results:

– big data analytics, deep learning, natural speech – mass collection of personal communication habits, surveillance

Time

• new methods → new insights
• brainstorming – ‘thinking outside the box’
• methodology in subfields of prosody

– more structure / pattern oriented, less function oriented

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

3

Time as a core concept Time functions, trajectories YARD (Yet Another Rhythm Discussion)

• annotation mining
• multiple oscillators
• low frequency spectra:
• amplitude modulation
• amplitude demodulation

Metatheoretical framework for time discussion Discussion of different paradigms Challenges for the future Evolution TIME

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

4

Strawson’s acoustic Gedankenexperiment:

• Can one conceive of an individual, i.e. an object or a person, in a

spaceless world with a time dimension only?

Conclusion:

• if changing sounds are assumed to have moving sources
• then the sources are interpretable as individual entities

Following this basic ontology:

• segments of speech signals are dynamically changing events
• not static units: speech has changing frequencies and rhythms

(though in practice it is often convenient to forget this)

Strawson, Peter F. Individuals. An Essay in Descriptive Metaphysics. London: Methuen. 1959.

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

5

In memoriam This work is dedicated to the memory of Wiktor Jassem, emeritus

• f the Polish Academy of Science, distinguished phonetician and

prosodist, pioneer in spectral analysis and speech synthesis, and authority on Polish phonetics and phonology, mentor and long-time friend, with whom I discussed the seeds of many ideas reflected in this presentation over some 30 years, in particular the use of the difference spectra discussed in the present talk. I would especially like to remember Grzegorz Dogil, formerly of Lublin, Poznań, Bielefeld and Stuttgart, whom many of you have known personally. Greg passed away much too soon six months

• ago. His highly productive early work in my department in Bielefeld,

including points related to the content of this address, was a source of inspiration for us, and has inspired many more phoneticians since that time.

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

6

Acknowledgments

to the strong tradition of logical and computational phonetic and phonological research in Poland, particularly in Poznań:

• 1. Wiktor Jassem, on acoustic edge detection and on hierarchies of intonation

and rhythm

In many publications

• 2. Tadeusz Batóg, on hierarchical mereological models in phonology

Batóg, Tadeusz. The Axiomatic Method in Phonology. London: Routledge and Kegan Paul, 1967.

• 3. Maria Steffen-Batóg finite state computational phonology

Steffen-Batóg, Maria. The problem of automatic phonemic transcription of written Polish. Biuletyn Fonograficzny. 14, pp. 75–86, 1973.

And to Batóg & Steffen-Batóg on formalizing phonetic distance

Steffen-Batóg, Maria and Tadeusz Batóg. A distance function in phonetics. Lingua Posnaniensis, XXIII, 47–58. 1980.

Special thanks for valuable hints, comments, suggestions and data:

Petra Wagner, Plinio Barbosa, Rosemarie Tracy, Alexandra Gibbon Yu Jue, Liang Jie, Liu Huangmei, Chen Wenjun (Shanghai) Lin Xuewei, Li Peng, He Linfang, Feng Baoyin, Bi Dan (Guangzhou)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

7

Thinking outside the box:

– different methods and method combinations – cooperation with other disciplines

• computational linguistics
• biology, ethology, musicology

– computational perspectives

• user
• script / software developer

– ((exploration* confirmation*)* standardization*)* cycles – exploratory rather than confirmatory research

• sometimes: lots of statistics and few examples
• here: lots of examples and not so much statistics
• discovery by thinking outside the box, visualisation,

analogy, ... TIME

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

8

Alternative Discourse Prosody

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

9

Time Types Major Time Domains Processing Time Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Patterns: static & dynamic Time Stamps: 3D Time Trees AM: Multiple Oscillators, Production emulation Rhythm is AM & FM Spectral Zones AM: Spectral Zones, Perception emulation FM: Discourse Modulation

Time: Types, Domains, Processes, Patterns, Stamps, Modulations Heuristic methods of prosodic pattern analysis:

• effects on subphones, subphone elements in sequence
• duration patterns of vocalic, syllabic, stress units
• pitch patterns at phone, syllable, word, … discourse rank

The background:

• epochs of time from discourse to evolution
• domains of time from syllable to discourse
• paradigmatic and syntagmatic patterns in time

Explanatory methods of prosodic pattern analysis:

• rhythm as similarity, isochrony and alternation
• rhythm as oscillation and modulation in speech production
• rhythm as demodulation of oscillation in speech perception

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

10

Time Types Major Time Domains Processing Time Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Patterns: static & dynamic Time Stamps: 3D Time Trees AM: Multiple Oscillators, Production emulation Rhythm is AM & FM Spectral Zones AM: Spectral Zones, Perception emulation FM: Discourse Modulation

Time: Types, Domains, Processes, Patterns, Stamps, Modulations Heuristic methods of prosodic pattern analysis:

• effects on subphones, subphone elements in sequence
• duration patterns of vocalic, syllabic, stress units
• pitch patterns at phone, syllable, word, … discourse rank

Explanatory methods of prosodic pattern analysis:

• rhythm as similarity, isochrony and alternation
• rhythm as oscillation and modulation in speech production
• rhythm as demodulation of oscillation in speech perception

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

11

Time Types Major Time Domains Processing Time Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Patterns: static & dynamic Time Stamps: 3D Time Trees AM: Multiple Oscillators, Production emulation Rhythm is AM & FM Spectral Zones AM: Spectral Zones, Perception emulation FM: Discourse Modulation

Time: Types, Domains, Processes, Patterns, Stamps, Modulations Explanatory methods of prosodic pattern analysis:

• rhythm as similarity, isochrony and alternation
• rhythm as oscillation and modulation in speech production
• rhythm as demodulation of oscillation in speech perception

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

12

Time: Types, Domains, Processes, Patterns, Stamps, Modulations

Time Types Major Time Domains Processing Time Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Patterns: static & dynamic Time Stamps: 3D Time Trees AM: Multiple Oscillators, Production emulation Rhythm is AM & FM Spectral Zones AM: Spectral Zones, Perception emulation FM: Discourse Modulation

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

13

Time and Prosody: overview Events vs. objects Five Major Time Epochs Four Time Types:

• Categorial time: paradigmatic
• Rubber time:

syntagmatic

• Clock time:

time stamps

• Cloud time:

real time Two kinds of processing time:

• recursion, time and space:
• finite vs. non-finite memory (space)
• linear vs. non-linear complexity (time)

Time Types Major Time Domains Processing Time Time Patterns: static & dynamic

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

14

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Various phonologies, but most explicitly Event Phonology

Time Types Major Time Domains Process Time Time Patterns

Thanks to Andras Kornai, for the concept ‘Rubber Time’

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

15

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Clock Time Cloud Time Various phonologies, but most explicitly Event Phonology Speech Technology ‘front ends’:

• Recognition
• Synthesis
• Identification

Thanks to Andras Kornai, for the concepts ‘Rubber Time’ and ‘Clock Time’

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

16

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Clock Time Cloud Time Various phonologies, but most explicitly Event Phonology Speech Technology ‘front ends’:

• Recognition
• Synthesis
• Identification

Word and sentence recognition systems, Text-to- Speech systems Thanks to Andras Kornai, for the concepts ‘Rubber Time’ and ‘Clock Time’

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

17

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Clock Time Cloud Time Various phonologies, but most explicitly Event Phonology Speech Technology ‘front ends’:

• Recognition
• Synthesis
• Identification

Time Map Phonology Word and sentence recognition systems, Text-to- Speech systems Thanks to Andras Kornai, for the concepts ‘Rubber Time’ and ‘Clock Time’

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

18

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Clock Time Cloud Time Various phonologies, but most explicitly Event Phonology Speech Technology ‘front ends’:

• Recognition
• Synthesis
• Identification

Time Map Phonology Word and sentence recognition systems, Text-to- Speech systems Thanks to Andras Kornai, for the concepts ‘Rubber Time’ and ‘Clock Time’

Time Types Major Time Domains Process Time Time Patterns

Categorial Time and ‘Rubber’ Time are, strictly speaking, metaphorical terms, and actually refer to abstract paradigmatic and syntagmatic structural relations. Time in the strict senses of Clock time and Cloud Time is not within the domain of phonology. (cf. Zhang’s critique in the Proceedings)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

19

1.Utterance in discourse:

– Milliseconds Micromotor activity: speech sounds – Seconds, minutes: Prosody

2.Individual language development:

– Years: Acquisition and learning

3.Social language change

– Pragmatic effects of language and speech ‘influencers’

4.Historical language & culture change - ‘dreamtime’

– Millennia: typological change, loss of inter-comprehensibility

5.Evolution:

– Multimillennia: differentiation of species communication

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

20

Processing time: a remark on recursion from a computational linguistic point of view In the many discussions of recursion over the past 20 years or so, a crucial distinction which affects processing time has been neglected:

– linear recursion: left & right branching (computationally

equivalent to iteration), iteration, with finite working memory and linear processing time (a function of the length of the input)

– non-linear

recursion: centre-embedding, cross-serial dependencies with unrestricted memory and at least quadratic processing time

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

21

Processing time: a remark on recursion from a computational linguistic point of view Food for thought:

– Arbitrary finite depth hierarchies cf. Phonological Hierarchy,

syllable phonotactics) also have linear processing time: realistic for speech

– Linear (right or left) recursion conditions: realistic for speech – Non-linear (centre-embedding) recursion: unrealistic for

speech

without a finite depth condition – though extra depth may be

made available through time and memory enhancement by means

• f rehearsed speech and writing) is available

Time Types Major Time Domains Process Time Time Patterns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

22

Categorial Time (paradigmatic relations) ‘Rubber’ Time (syntagmatic relations) Clock Time Cloud Time Various phonologies, but most explicitly Event Phonology Speech Technology ‘front ends’:

• Recognition
• Synthesis
• Identification

Time Map Phonology Word and sentence recognition systems, Text-to- Speech systems Thanks to Andras Kornai, for the concepts ‘Rubber Time’ and ‘Clock Time’

Time Types Major Time Domains Process Time Time Patterns

Time and Prosody: summary

• Events vs. objects
• Five Major Time Epochs
• Four Time Types:
• Categorial time: paradigmatic
• Rubber time:

syntagmatic

• Clock time:

time stamps

• Cloud time:

real time

• Two kinds of processing time:
• recursion, time and space:
• finite vs. non-finite memory
• linear vs. non-linear complexity

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

23

Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Stamps: 3D Time Trees Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Stamps: 3D Time Trees

Annotation with time stamps: overview

• Heuristic approaches
• rhythm: the truth – but not the whole truth
• Annotation: event property + time stamps
• Annotation mining: information extraction from annotations
• Rhythm definition:

similarity + isochrony + alternation

• 1D dispersion measures: duration variability
• 2D area measures: duration quadrant
• 3D hierarchical analysis:
• Time Tree Analysis – induction of duration graphs
• application of TTA to annotation mining

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

24

Signal Annotation:

• segmentation
• classification

Data Repository DSP Hardware Software

Manual calculation LOcalc Excel SPSS Stata MatLab R Python Praat Speech engineering software development

Time Stamps

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

Analysis

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

25

1-dimensional time-stamp duration analysis:

• means of

sequences (Var, PIM, PFD) – no compensation from tempo change pairs (PVI) – abstracts away from tempo change

• no account of rhythm as an alternation relation, and only binary relations

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

26

Wagner, Petra (2007). “Visualizing levels of rhythmic organisation.” Proc. International Congress of Phonetic Sciences, Saarbrücken 2007, pp. 1113-1116, 2007

2-dimensional time-stamp duration analysis:

• classification of alternation relations in z-scored scatter plot

Mandarin: means scattered relatively evenly around the centre English: e.g. count(short-short) > count(long-long)

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

LONG- LONG LONG- SHORT SHORT- SHORT SHORT- LONG

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

27

Wagner, Petra (2007). “Visualizing levels of rhythmic organisation.” Proc. International Congress of Phonetic Sciences, Saarbrücken 2007, pp. 1113-1116, 2007

2-dimensional time-stamp duration analysis:

• classification of alternation relations in z-scored scatter plot

Mandarin: means scattered relatively evenly around the centre English: e.g. count(short-short) > count(long-long)

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

Mandarin Even clustering around the mean English Highly skewed: majority of short-short syllable relations, thus NOT BINARY

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

28

Gibbon, Dafydd. 2006. “Time types and time trees: Prosodic mining and alignment of temporally annotated data”. In: Stefan Sudhoff, et al., eds. Methods in Empirical Prosody Research. Berlin: Walter de Gruyter, pp. 281–209, 2006.

3-dimensional time-stamp duration analysis: time-tree induction:

• hierarchical classification of alternation relations
• several possible processing options: binary/nonbinary, lower/higher percolated
• related to phrasal and discourse patterns

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

29

Gibbon, Dafydd. 2006. “Time types and time trees: Prosodic mining and alignment of temporally annotated data”. In: Stefan Sudhoff, et al., eds. Methods in Empirical Prosody Research. Berlin: Walter de Gruyter, pp. 281–209, 2006.

3-dimensional time-stamp duration analysis: time-tree induction:

• hierarchical classification of alternation relations
• several possible processing options: binary/nonbinary, lower/higher percolated
• related to phrasal and discourse patterns

Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

Duration value upward percolation

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

30

Can be thought of as inverses of metrical generation algorithms (Compound and Nuclear Stress Rules) Inductive input-output relation (examples) Iambic (weak-strong) directionality, iNSR: ((miss . 3) (jones . 2) (came . 3) (home . 1)) → (r (w (w miss) (s jones)) (s (w came) (s home))) Trochaic (strong-weak) directionality, iCSR: ((light . 1) (house . 3) (keep . 2) (er . 3)) → ((r (s (s light) (w house)) (w (s keep) (w er))))

Gibbon, Dafydd. 2006. “Time types and time trees: Prosodic mining and alignment of temporally annotated data”. In: Stefan Sudhoff et al., eds. Methods in Empirical Prosody Research. Walter de Gruyter, pp. 281–209, 2006. Time Stamps

Annotation Mining

Time Stamps

1D Duration Dispersion Isochrony

Time Stamps

2D Duration Dispersion Scatter Plots

Time Stamps

3D Duration Dispersion Time Trees

parse trees, root at bottom

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

31

Annotation with time stamps: summary

• Heuristic approaches
• rhythm: the truth – but not the whole truth
• Annotation: event property + time stamps
• Annotation mining: information extraction from annotations
• Rhythm definition:

similarity + isochrony + alternation

• 1D dispersion measures: duration variability
• 2D area measures: duration quadrant
• 3D hierarchical analysis:
• Time Tree Analysis – induction of duration graphs
• application of TTA to annotation mining

Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Stamps: 3D Time Trees Time Stamps: Annotation Mining Time Stamps: 1D Isochrony Time Stamps: 2D Relations Time Stamps: 3D Time Trees

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

32

Help! I can’t see the wood for the trees!

So what is THE rhythm of a language? … Is this the right question?

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

33

Rhythm as iteration

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

34

So what is THE rhythm of a language? … Is this the right question?

• No. it is not the right question.

There are many rhythms in speech, of in many frequency ranges. Speech rhythms are unstable, ‘fuzzy’ hierarchies. Definitely not ‘quartz timing’. Because of this unstable property, rhythm types are spread

• ver rhythm zones, i.e. frequency zones. Approximately:
• phones:

20Hz … 10Hz (50ms … 100ms)

• syllables:

10Hz … 4Hz (100ms … 250ms)

• accents:

4Hz … 1Hz (250ms … 500ms)

• groups:

< 1Hz

• utterances: << 1Hz

Can we measure – or a least visualise these? Yes, we can. But we need to think in Hz, not ms.

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

35 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonological and Phonetic Oscillators: overview

• Explanatory models
• Phonological oscillators:
• iteration and linear recursion (left or right branching)
• iterative intonation models
• iterative tone sandhi models
• Phonetic oscillators:
• production models
• amplitude modulation frequencies (‘sonority’)
• frequency modulation frequencies (F0)
• perception models:
• amplitude envelope demodulation spectrum
• frequency demodulation models (pitch)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

36 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonological ‘oscillators’: overview

• Iteration as abstract oscillation
• Thus: iteration as the foundation of rhythm:
• English intonation (Pierrehumbert)
• Niger-Congo tone sandhi (Gibbon)
• Tianjin Mandarin sandhi (Jansche)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

37

Or as an equivalent regular expression: (( %H|%L ( H*|L*|H*+L|H+L*|L*+H|L+H* )+ H-|L- )+ H%|L% )+

Or as an equivalent right branching regular (type 3) grammar: IP → initb PiA PiA → pa PiA PiA → pa IntP IntP → interb PA IntP → interb IPend IPend → intonb IPend → intonb IP and vocabulary: initb : { H%, L% } interb : { H-, L- } intonb : { H%, L% } pa : { H*, L*, L*+H-, L-+H*, H*+L-, H-+L*, H*+H-} Pierrehumbert’s regular grammar as a finite state transition network

A phonological view of rhythm as iteration

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

38

Empirical overgeneration

1) Accents in a sequence tend to be all H* or all L* 2) Global contours tend to be rising with L* accents, falling with H* accents 3) Global contours may span more than 1 turn

Empirical undergeneration

1) Paratone hierarchy not included 2) No time constraints

A phonological view of rhythm as iteration

Pierrehumbert’s regular grammar as a finite state transition network

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

39

1-tape (1-level) transition network

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

40

2-tape (2-level) transition network

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

41

3-tape (3-level) transition network

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

42

Martin Jansche 1998 Tianjin Mandarin tone sandhi

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

43

Martin Jansche 1998 Tianjin Mandarin tone sandhi

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonological ‘oscillators’: summary

• Iteration as abstract oscillation
• Thus: iteration as the foundation of rhythm:
• English intonation (Pierrehumbert)
• Niger-Congo tone sandhi (Gibbon)
• Tianjin Mandarin sandhi (Jansche)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

44 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonetic Oscillators: overview

• Oscillations in emulations of speech production:
• coupled oscillators
• time-domain coupling (syllable ~ phrase)
• interlocutor entrainment
• Oscillation in emulations of speech perception:
• Amplitude vs. Frequency Modulation
• Amplitude demodulation
• waveform rectification and envelope extraction
• AM envelope spectrum with zone edge detection
• F0 demodulation (aka pitch trackiing)
• F0 spectrum with zone edge detection
• Hierarchical induction of spectral zones

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

45

INFORMATION

AMPLITUDE MODULATION CARRIER FREQUENCY NOISE FREQUENCIES + ✕ FILTER COEFFICIENTS INFORMATION FREQUENCY MODULATION

SPECTRAL ANALYSES in different frequency zones, COORDINATION AMPLITUDE DEMODULATION in different time zones rectification, LP filtering envelope detection FREQUENCY DEMODULATION in different frequency zones pitch tracking, formant tracking

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

46

Diode rectifier in a crystal set

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

47

Selected Work on Amplitude Envelope Demodulation Spectra

[1] Cummins, Fred, Felix Gers and Jürgen Schmidhuber. “Language identification from prosody without explicit features.” Proc. Eurospeech. 1999. [2] He, Lei and Volker Dellwo. “A Praat-Based Algorithm to Extract the Amplitude Envelope and Temporal Fine Structure Using the Hilbert Transform.” In: Proc. Interspeech 2016, San Francisco, pp. 530-534, 2016. [3] Hermansky, Hynek. “History of modulation spectrum in ASR.” Proc. ICASSP 2010. [4] Leong, Victoria and Usha Goswami. “Acoustic-Emergent Phonology in the Amplitude Envelope of Child-Directed Speech.” PLoS One 10(12), 2015. [5] Leong, Victoria, Michael A. Stone, Richard E. Turner, and Usha Goswami. “A role for amplitude modulation phase relationships in speech rhythm perception.” JAcSocAm, 2014. [6] Liss, Julie M., Sue LeGendre, and Andrew J. Lotto. “Discriminating Dysarthria Type From Envelope Modulation Spectra.” Journal of Speech, Language and Hearing Research 53(5):1246–1255, 2010. [7] Ludusan, Bogdan Antonio Origlia, Francesco Cutugno. “On the use of the rhythmogram for automatic syllabic prominence detection.” Proc. Interspeech, pp. 2413-2416, 2011. [8] Ojeda, Ariana, Ratree Wayland, and Andrew Lotto. “Speech rhythm classification using modulation spectra (EMS).” Poster presentation at the 3rd Annual Florida Psycholinguistics Meeting, 21.10.2017, U Florida. 2017. [9] Tilsen Samuel and Keith Johnson. “Low-frequency Fourier analysis of speech rhythm.” Journal of the Acoustical Society of America. 2008; 124(2):EL34–EL39. [PubMed: 18681499] [10] Tilsen, Samuel and Amalia Arvaniti. “Speech rhythm analysis with decomposition of the amplitude envelope: Characterizing rhythmic patterns within and across languages.” The Journal of the Acoustical Society of America 134, p. 628 .2013. [11] Todd, Neil P. McAngus and Guy J. Brown. “A computational model of prosody perception.” Proc. ICSLP 94, pp. 127-130, 1994. [12] Varnet, Léo, Maria Clemencia Ortiz-Barajas, Ramón Guevara Erra, Judit Gervain, and Christian Lorenzi. “A cross-linguistic study of speech modulation spectra.” JAcSocAm 142 (4), 1976–1989, 2017. AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

☛ ☛ ☛

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

48

Amplitude Envelope Modulation Spectrum (AEMS, AMS, EMS) Frequency Zones

Amplitude Envelope Modulation

↓

Amplitude Envelope Demodulation

absolute value of Hilbert transform (or rectification & peak-picking / LP filtering)

↓

Spectral slice (FFT)

↓

Spectral Zone Edge Detection

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

49 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

50

Rectified modulated signal (light green, top) Signal: 2s, 200×5 Hz AM carrier (light & dark green) Demodula ted FM (‘pitch’) track (red

• utline)

AM and FM spectra

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

AM and FM spectra as heatmaps Frequency Zone Edge Detection Demodula ted AM envelope (red

• utline)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

51 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

52

English (RP) Edinburgh corpus “The North Wind and the Sun” Beijing Mandarin Yu corpus “bei3 feng1 gen1 tai4 yang2”

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

53

English (RP) Edinburgh corpus “The North Wind and the Sun” Beijing Mandarin Yu corpus “bei3 feng1 gen1 tai4 yang2” Short phrases Short IPUs Paraton e IPUs IPU hierarchy Phrases IPUs

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

1 Hz

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

54

Spectral Frequency Zone Boundaries English Newsreading

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

English Story

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

55

Spectral Frequency Zone Boundaries

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

56

L- strong, > L- strong, < R- strong, > R- strong, <

Frequency Trees: Spectral Zone Hierarchies

English Newsreading

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

57

L- strong, < AEMS Frequency Tree English Newsreading

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

58

L- strong, > L- strong, < R- strong, > R- strong, < English North Wind & Sun

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Frequency Trees: Spectral Zone Hierarchies

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

59

L- strong, < AEMS Frequency Tree English North Wind & Sun

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

60

L- strong, > L- strong, < R- strong, > R- strong, < Mandarin North Wind & Sun

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Frequency Trees: Spectral Zone Hierarchies

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

61

L- strong, < AEMS Frequency Tree Mandarin North Wind & Sun

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

62 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Next step: Distance analysis of AEMS of 5s adjacent audio clips, English & Mandarin Next but one step: Conventional analysis of AEMS edges in 5 second audio clips, English & Mandarin

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

63 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Data:

“The North Wind and the Sun” Male, English: 40s Female, Mandarin: 40s

Method:

Comparison of non-overlapping adjacent 5s audio chunks

– offsets into recording: 0, 5, 10, 15, 20, 25, 30, 35 – AEMS for each chunk – Inter-speaker comparison (AEMS pointwise means, r=0.82) – Comparison by hierarchical similarity / distance

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

64 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Similarity criterion: >3 adjacent same-speaker settings Largest per speaker score: (4+3)/16 Largest cluster: 4/16 Phonetic distance between consecutive chunks of AEM spectra Task: compare 7 hierarchical clustering algorithms

• cf. pioneering work on phonetic distance by Steffen-Batóg

related to methods used in stylometry dialectometry typological language classification

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

65 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Highest speaker- specific total:

1 Nrst.Pt. (4+3+3)/10

Largest cluster:

5 UPGMC 5/10

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

66 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonetic Oscillators: summary

• Oscillations in emulations of speech production:
• coupled oscillators
• time-domain coupling (syllable ~ phrase)
• interlocutor entrainment
• Oscillation in emulations of speech perception:
• Amplitude vs. Frequency Modulation
• Amplitude demodulation
• AEMS and AEMDS edge detection
• F0 demodulation (aka pitch trackiing)
• F0 spectrum with zone edge detection
• Hierarchical induction of spectral zones

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

67 AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Phonological and Phonetic Oscillators: summary

• Explanatory models
• Phonological oscillators:
• iteration and linear recursion (left or right branching)
• iterative intonation models
• iterative tone sandhi models
• Phonetic oscillators:
• production models
• amplitude modulation frequencies (‘sonority’)
• frequency modulation frequencies (F0)
• perception models:
• amplitude envelope demodulation spectrum
• frequency demodulation models (pitch)

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

68

AM: Multiple Oscillators, Production emulation Rhythm is AM & FM Spectral Zones AM: Spectral Zones, Perception emulation FM: Discourse Modulation

FM and Discourse Modulation: overview

• Rank Interpretation Architecture for spoken language
• Prosody and discourse:
• AM and FM spectra - correlations
• Constraints on pitch accent sequences
• Syntagmatic entrainment in adjacency pairs
• Emotive exclamations and whistles

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

69

People and Signs Denotation, Reference Cloud Time semiotic relation Categorial Time simple and structured forms Modality Interpretation hierarchical patterning Focus Contrast Emphasis

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Ternary semiotic basis for signs at all ranks

linear patter ns

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

70

Discourse: Monologue, Dialogue Utterance: turn, IPU, ... Sentence, clause, phrase Word: simple, inflected, compound, derived

Rank Interpretation Architecture

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

71

Discourse prosody, Case 1: AM vs. FM spectra If a spectrum can be derived from the AM envelope, why not derive a spectrum from the FM track and see whether they correlate?

Preliminary answer: Yes, they do correlate, but not overwhelmingly strongly, and depending on which subspectra are measured.

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

72

Mandarin, female 30 sec, < 20Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

73

Mandarin, female 30 sec, < 5Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

74

Mandarin, female 30 sec, < 1Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

75

English, male 30 sec, < 20Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

76

English, male 30 sec, < 5Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

77

English, male 30 sec, < 1Hz

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

78

Discourse prosody, Case 2: Accent constraints

Constraint 1:

Pitch accents in the same sequence tend to be of the same type and collocate with specific global contours

Constraint 2:

Pitch accent sequences tend to match the final phrasal accent:

– low rising types tend to be followed by a rising final accent – high rising types tend to be followed by a rising final accent

Constraint 3:

Pitch accent sequence types tend to match information structure and

– low pitch accent sequences tend to be introductory or questioning – high pitch accent sequences tend to be closing or stating

with typologically relevant constraint violations in different languages and dialects

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

79

Answer: falling utterance contour L* sequence, global rise, final rise H* sequence, global fall H* sequence, global fall L* sequence, global fall, final rise Response Continuati

• n

Interview start Question Questi

• n

Cantonese area (Guangzhou) Cantonese area (Guangzhou)

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

80

Discourse prosody, Case 3: Long FM contours

Thesis: in evolution,

– frequency modulation and rhythm came first

• emotional cries
• turn-taking came before grammar,

Levinson, “Turn-taking in Human Communication – Origins and Implications for Language Processing”, 2015

Note: in infant speech,

– frequency modulation and rhythm also come first

• emotional cries

Wermke, Sebastian-Galles

• turn-taking
• cf. the ‘bootstrapping’ literature

the infant ‘twin-talk’ videos on YouTube ☺

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

81

Answer: falling utterance contour Question+Answer: rising-falling adjacency pair contour

syntagmatic entrainment

Question: rising utterance contour

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

82

Discourse Prosody, Case 4: emotive FM contours

Thesis 1:

In the evolutionary time domain:

emotive modulations came before structural modulations

Thesis 2:

In the beginning was “Wow!” (Or “Aaah!”)

Thesis 3:

Or the wolf whistle (it’s not simply ‘cat-calling’)

Thesis 4:

In any case, other primates wowed, aahed and whistled first – we continued the custom

Is this why in some societies whistling is tabooed?

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

83

哇

Cantonese region (Guangzhou) Wu region (Shanghai) EMOTIVE EXCLAMATIONS

‘Tone 6’ ☺

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

Tone 4

Twin peaks: 2nd formant + pitch

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

84

啊

Cantonese region (Shenzhen)

EMOTIVE EXCLAMATIONS

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

85

TELEGLOSSIA

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

86

Street whistle Cantonese shoolboy Primate coloratura soprano

TELEGLOSSIA

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

87

Street whistle Cantonese shoolboy In fact, it’s a black-handed gibbon

TELEGLOSSIA

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

88

Cantonese street whistles (child, middle school) Black-handed gibbon calls

TELEGLOSSIA

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

89

Street whistle Cantonese shoolboy In fact, it’s a black-handed gibbon

TELEGLOSSIA

AM

Multiple Oscillators, production emulation

Rhythm

is

• scillation

and iteration

AM

Spectral Zones, perception emulation

FM

discourse turns emotion

FM and Discourse Modulation: summary

• Rank Interpretation Architecture for spoken language
• Prosody and discourse:
• AM and FM spectra - correlations
• Constraints on pitch accent sequences
• Syntagmatic entrainment in adjacency pairs
• Emotive exclamations and whistles

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

90

… thinking outside the box Summary: Conclusion:

SP9, Poznań, 13 June 2018

• D. Gibbon, The Future of Prosody - It's about Time

91

Thank you! 谢谢 ! Dziękuję!