Byron T. Ahn | Intonational Meaning + Variation

This int² project is exploring details of how intonational forms and semantic/pragmatic meanings map onto one another.

Below is the abstract from the NSF grant:

Exploring Variation in English Intonational Acoustic Phonetics from Grammatical Perspectives

Spoken language consists of multiple streams of information – including a segmental stream (for example, the consonants and vowels that make up words, suffixes, and prefixes) and a suprasegmental stream (such as the pitch, amplitude, and timing that define stress and cadence) which is often called “prosody”. All speakers (consciously or not) know prosody can be used to communicate meaning – a single sequence of words can be pronounced many ways, often with importantly distinct meanings. Despite this, most formal and computational linguistic investigations into meaning have primarily dealt with meanings encoded by the segmental stream (in words, suffixes, and prefixes). This project addresses this issue by exploring questions of how prosody and meaning relate. For example, what sorts of prosodic changes map onto changes in meaning? And what sorts of meanings? To explore these questions, the research team makes use of contemporary empirical methodologies, analytic tools, and formal theories, and particular attention is paid to intonational aspects of prosody.

The project encompasses three sets of experimental tasks of speech production and speech perception. Each of these sets of tasks corresponds to a different domain of semantic/pragmatic meanings: Common Ground management, Information Structure signalling, and scope ambiguity resolution. Different semantic/pragmatic contexts correspond to different experimental conditions, and these contexts are presented to participants as visual, comic strip-like, stimuli. These experiments measure prosodic characteristics of speech (e.g., pitch movements, coded with a recently developed annotation system called PoLaR) and/or reactions to speech (e.g., acceptability ratings). These data are then subjected to quantitative analysis as well as to formal linguistic models to uncover how different meaning-based conditions and different prosodic patterns are related. This work has important implications for improving human-computer interactions that are mediated by natural language, as well as for deepening understanding of various speech pathologies.

This research will be investigated through a series of experiments, following a general pattern laid out in the flowchart below:

A flowchart laying out the paradigm for expreimental research on intonational meaning

We envision three experimental sets, briefly named below:

Experiment Set 1: Common Ground Management and Intonational Meaning
Experiment Set 2: Information Structure and Phonemic Inventory of Mainstream U.S. English
Experiment Set 3: Conventionalized Intonational Meaning and Scope Ambiguity

Supplementing this, we begin by applying the analytical methods we will use to a dataset previously collected by Rett & Sturman (2021) to explore the prosody of exclamative utterances in English.

some work in this project

Ahn, Byron, Nanette Veilleux, Beth Sturman, Alejna Brugos, Sunwoo Jeong & Stefanie Shattuck-Hufnagel. 2022, March. How Meaningful These Intonational Contours Are! Poster presented at 35th Annual Conference on Human Sentence Processing (formerly the ‘CUNY Conference on Human Sentence Processing’). University of California, Santa Cruz.
abstract Slides Bib

As noted in works like Rett and Sturman 2021, exclamative expressions in English can take on a variety of specialized syntactic forms, which have been argued to share a syntactic core (responsible for word-order effects) and a semantic core (a mirativity operator). Like Rett and Sturman, this work aims to identify an intonational core for exclamatives, and successfully replicates their results using an entirely different methodology: a non-phonological annotation system (PoLaR) alongside statistical analysis and machine learning techniques facilitated by those labels.

Constructing a grammatical analysis this way has advantages (including practical ones like PoLaR’s ease of use). Moreover, developing these models can serve as a stepping stone towards computationally identifying more complex models of other sem-prag and/or paralinguistic meanings.
@unpublished{Ahn-Veilleux-Sturman-Brugos-Jeong-Shattuck-Hufnagel-HSP-exclamatives, author = {Ahn, Byron and Veilleux, Nanette and Sturman, Beth and Brugos, Alejna and Jeong, Sunwoo and Shattuck-Hufnagel, Stefanie}, howpublished = {Poster presented at 35th Annual Conference on Human Sentence Processing (formerly the ‘CUNY Conference on Human Sentence Processing’). University of California, Santa Cruz}, title = {How Meaningful These Intonational Contours Are!}, year = {2022} }
Ahn, Byron, Stefanie Shattuck-Hufnagel & Nanette Veilleux. 2016. Evidence and Intonational Contours: An Experimental Approach to Meaning in Intonation. Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology 189–192.
abstract Paper Bib

One proposed function of prosody is conveying speakers’ stances on the evidence for their statements and assessments of listeners’ beliefs. Testing this is challenging - specifying evidential status is difficult, and speakers vary intonationally for a given context. A novel task of reading lines from comic strips elicits relatively consistent intonation, suggesting both the method’s usefulness, and the efficacy of speaker beliefs in governing prosodic contours. Preliminary results suggest H* accents are used when speakers believe they have evidence the hearer lacks, and L* accents for the flipped situation.
@article{Ahn-Shattuck-Hufnagel-Veilleux-SST2016, author = {Ahn, Byron and Shattuck-Hufnagel, Stefanie and Veilleux, Nanette}, journal = {Proceedings of the Sixteenth {A}ustralasian International Conference on {S}peech {S}cience and {T}echnology}, pages = {189--192}, title = {Evidence and Intonational Contours: An Experimental Approach to Meaning in Intonation}, year = {2016} }