Post-nasal devoicing and the blurring process

Beguš, Gašper

doi:10.1017/s002222671800049x

Cited by 50 publications

(66 citation statements)

References 112 publications

(325 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In other words, we model phonological computation on a cognitive level as if no articulatory constraints were present in human speech. This is highly desired for the task of distinguishing those aspects of phonology that are influenced by cognitive factors from those that are influenced by articulation, motor planning, or historical developments (Beguš, 2018a).…”

Section: Future Directionsmentioning

confidence: 99%

Generative Adversarial Phonology: Modeling unsupervised allophonic learning with neural networks

Beguš¹

2019

Preprint

View full text Add to dashboard Cite

Training deep neural networks on well-understood dependencies in speech data can provide new insights into how they learn internal representations. This paper argues that acquisition of speech can be modeled as a dependency between random space and generated speech data in the Generative Adversarial Network architecture and proposes a methodology to uncover the network's internal representations that correspond to phonetic and phonological properties. The Generative Adversarial architecture is uniquely appropriate for modeling phonetic and phonological learning because the network is trained on unannotated raw acoustic data and learning is unsupervised without any language-specific assumptions or pre-assumed levels of abstraction. A Generative Adversarial Network was trained on an allophonic distribution in English, in which voiceless stops surface as aspirated word-initially before stressed vowels, except if preceded by a sibilant [s]. The network successfully learns the allophonic alternation: the network's generated speech signal contains the conditional distribution of aspiration duration. The paper proposes a technique for establishing the network's internal representations that identifies latent variables that correspond to, for example, presence of [s] and its spectral properties. By manipulating these variables, we actively control the presence of [s] and its frication amplitude in the generated outputs. This suggests that the network learns to use latent variables as an approximation of phonetic and phonological representations. Crucially, we observe that the dependencies learned in training extend beyond the training interval, which allows for additional exploration of learning representations. The paper also discusses how the network's architecture and innovative outputs resemble and differ from linguistic behavior in language acquisition, speech disorders, and speech errors, and how well-understood dependencies in speech data can help us interpret how neural networks learn their representations.

show abstract

Section: Future Directionsmentioning

confidence: 99%

Generative Adversarial Phonology: Modeling unsupervised allophonic learning with neural networks

Beguš¹

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Despite these objections, mechanisms exist within the channel bias approach to derive typology that go further than the simple statement that rare sound changes produce rare alternations. Based on a typological study of an unnatural process, postnasal devoicing, Beguš (2019) argues that unnatural processes require at least three sound changes (as opposed to at least two for unmotivated processes and at least one for natural processes; the M inimal S ound C hange R equirement ), which explains the relative rarity of processes with different degrees of naturalness. The idea that unmotivated processes are rare because they require a complex history is certainly not new (Bell 1970, 1971, Greenberg 1978: 75–76, Cathcart 2015, Morley 2015), but the Minimal Sound Change Requirement explains why unnatural processes are the least frequent (compared to natural or unmotivated processes; see §2).…”

Section: Introductionmentioning

confidence: 99%

“…Due to its design, however, Cathcart's (2015) model relies on the representativeness of diachronic surveys for all sound changes, not only for those that are estimated (see also §3.2), and is computationally demanding, making it difficult to implement. The models in Greenberg (1978) and Cathcart (2015) also do not take into consideration the crucial distinctions made in Beguš (2019: 744): ‘the subdivision of unusual rules into unnatural versus unmotivated rules, paired with the proof that the latter require at least three sound changes to arise’ 2 . The model proposed in this paper has the disadvantage that the trajectories of sound changes that lead to a certain alternation need to be identified manually (similar to the Bell and Greenberg models), but this also means that samples of sound changes need be representative only for the sound changes being estimated.…”

Section: Introductionmentioning

confidence: 99%

“…/p/ → [b] / m ＿) and postnasal devoicing (/b/ → [p] / m ＿), intervocalic voicing (/p/ → [b] / V＿V) and intervocalic devoicing (/b/ → [p] / V＿V), and final devoicing (/b/ → [p] / ＿#) and final voicing (/p/ → [b] / ＿#). The feature [±voice] is chosen because phonetic naturalness is probably best understood precisely for this feature (see Ohala 1983, 2011 and Westbury & Keating 1986 on the ‘Aerodynamic Voicing Constraint’, and for detailed argumentation Beguš 2019), and all alternations have been well researched typologically. The alternations also differ in their degree of synchronic attestedness (see Fig.…”

Section: Introductionmentioning

confidence: 99%

“…This paper makes use of several diachronic concepts from Beguš (2019). First, it adopts the division of phonological processes into natural , unmotivated and unnatural .…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Estimating historical probabilities of natural and unnatural processes

Beguš

2020

Phonology

Self Cite

View full text Add to dashboard Cite

This paper presents a technique for estimating the influences of channel bias on phonological typology. The technique, based on statistical bootstrapping, enables the estimation of historical probability, the probability that a synchronic alternation arises based on two diachronic factors: the number of sound changes required for an alternation to arise and their respective probabilities. I estimate historical probabilities of six attested and unattested alternations targeting the feature [voice], compare historical probabilities of these alternations, perform inferential statistics on the comparison and, to evaluate the performance of the channel bias approach, compare outputs of the diachronic model against the independently observed synchronic typology. The technique also identifies mismatches between the typological predictions of the analytic bias and channel bias approaches. By comparing these mismatches with the observed typology, this paper attempts to quantitatively evaluate the distinct contributions of the two influences on typology in a set of alternations targeting the feature [voice].

show abstract

The blurring history of intervocalic devoicing

Beguš,

Dąbkowski

2024

J. Ling.

View full text Add to dashboard Cite

The intervocalic position favors voicing in stops. Yet, some languages have been reported to feature the opposite (unnatural) process of intervocalic devoicing. This paper investigates two such case studies. Pre-Berawan intervocalic *b and *g have developed into Berawan k. Pre-Kiput intervocalic *g, *ɟʝ, and *v have developed into Kiput k, cç, and f, respectively. To account for the data, we invoke Beguš’s (2018, 2019) blurring process model of sound change. The model proposes that unnatural phonology derives from a sequence of at least three phonetically motivated sound changes. We argue that the steps involved in intervocalic devoicing are (i) the intervocalic fricativization of voiced stops, (ii) devoicing of fricatives, and (iii) the occlusion of devoiced fricatives. Each of the steps is independently attested and motivated. We demonstrate that our blurring process proposal explains aspects of the historical development unaccounted for by previous approaches, and present new evidence suggesting that a single sound change could not have operated in the prehistory of Berawan. Thus, we maintain the conservative position that unnatural diachronic developments arise from sequences of natural and phonetically grounded sound changes.

show abstract

Post-nasal devoicing and the blurring process

Cited by 50 publications

References 112 publications

Generative Adversarial Phonology: Modeling unsupervised allophonic learning with neural networks

Generative Adversarial Phonology: Modeling unsupervised allophonic learning with neural networks

Estimating historical probabilities of natural and unnatural processes

The blurring history of intervocalic devoicing

Contact Info

Product

Resources

About