A Cognitively Grounded Measure of Pronunciation Distance

Wieling, Martijn; Nerbonne, John; Bloem, Jelke; Gooskens, Charlotte; Heeringa, Wilbert; Baayen, R. Harald

doi:10.1371/journal.pone.0075734

Cited by 20 publications

(11 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The often arbitrary decisions that have to be made here to get neighborhood measures to work are not necessary in NDL. Putting this technical problem aside, it is worth noting that computationally, NDL activations and a computational measure for phonetic string similarity, a weighted edit distance, as developed by Wieling et al (2012) are functionally equivalent (Wieling et al, 2014). Crucially, the weights used in this edit distance quantify the functional load of an edit across the vocabulary, which is the functional equivalent of what NDL networks accomplish using the Rescorla-Wagner learning rule.…”

Section: Discussionmentioning

confidence: 99%

Opposing forces on acoustic duration

Tucker¹,

Sims²,

Baayen³

2019

Preprint

View full text Add to dashboard Cite

The present paper investigates the influence of opposing lexical forces on speech production using the duration of the stem vowel of regular and irregular verbs as attested in the Buckeye corpus of conversational North-American English. We compared two sets of predictors, reflecting two different approaches to speechproduction, one based on competition between word forms, the other based on principles of discrimination learning. Classical measures in word form competition theories such as word frequency, lexical density, and gang size (types of vocalic alternation) were predictive of stem vowel duration. However, more precise predic-tions were obtained using measures derived from a two-layer network model trained on the Buckeye corpus. Measures representing strong bottom-up support predicted longer vowel durations. Conversely, measures reflecting uncertainty predicted shorter vowel durations, including a measure of the verb’s semantic density. The learning-based model also suggests that it is not a verb’s frequency as such that gives rise to shorter vowel duration, but rather a verb’s collocational diversity. Results are discussed with reference to the Smooth Signal Redundancy Hypothesis and the Paradigmatic Signal Enhancement Hypothesis.

show abstract

Section: Discussionmentioning

confidence: 99%

Opposing forces on acoustic duration

Tucker¹,

Sims²,

Baayen³

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…We considered place, manner and voice as attributes of consonants, and frontness, openness, diphthong and tone/length (only when used for contrasting meanings) as attributes of vowels. This implementation was motivated by the fact that Levenshtein distance does consider the degrees of similarity existing between phonemes from a perceptual point of view: for example, /p/ and /b/ are more easily confusable than /p/ and /f/, and therefore a substitution of the first kind should account as smaller (Heeringa, 2004;Wieling et al, 2014). It also reduces the effects of subjectivity in transcription: inaccuracies in easily confusable phonemes, which are more likely to happen, have a relatively small effect.…”

Section: Transmission Errormentioning

confidence: 99%

Modality effects in the cultural evolution of language: An experimental iterated learning approach

Perelman

2019

Onomázein

View full text Add to dashboard Cite

Journal of linguistics, philology and translation ONOMÁZEIN 45 (September 2019): 103 -125 Fernanda Weinstein Perelman Modality effects in the cultural evolution of language: An experimental iterated learning approach 104 Cultural evolution has been proposed as the mechanism by which human languages' distinct features emerge. One of such features is structure, which is regarded as an optimal solution to the competing pressures for simplicity and expressivity in language learning and use. A recent experimental iterated learning study shows that structure can emerge from an unstructured language under these competing pressures, by implementing both a learning and a communication task in a transmission chain setup. However, as most iterated learning experiments, it was run on a written modality, which might be problematic if the aim is to drive conclusions about language in general-writing is not language's default modality and was not present in early stages of language evolution.The present study carried out a partial replication of the aforementioned experiment, contrasting a written condition (analogous to the original) with a spoken condition, in order to test for possible modality effects. Results for the written condition did not replicate those in Kirby et al. (2015) in any of the measures, suggesting that motivational factors could have played a crucial role in the previous findings. This hinders the interpretations of modality effects and suggests the need of further work. Abstract

show abstract

“…Wieling et al (2014a) analyze foreign accents using a refinement of the edit-distance measure developed in dialectometry [i.e., using sensitive sound segment distances obtained via pointwise mutual information (PMI); Wieling et al , 2012 and validate the measure using native speakers' judgments of native-like versus foreign sounding, thereby obtaining a strong correlation (r ¼ 0.8) for a logarithmically transformed version of the refined edit-distance measure. Wieling et al (2014c) compare this measure with a completely different computational measure (i.e., one based on the theory of human discriminative learning) and show that the two methods achieve comparable performance. In effect, this observation suggests that when comparing pronunciations at an aggregate level by averaging the pronunciation distances of many words, the specific measure used might not be so important.…”

Section: Related Workmentioning

confidence: 99%

Advances in Dialectometry

Wieling

Nerbonne

2015

Annu. Rev. Linguist.

Self Cite

View full text Add to dashboard Cite

Dialectometry applies computational and statistical analyses within dialectology, making work more easily replicable and understandable. This survey article first reviews the field briefly in order to focus on developments in the past five years. Dialectometry no longer focuses exclusively on aggregate analyses, but rather deploys various techniques to identify representative and distinctive features with respect to areal classifications. Analyses proceeding explicitly from geostatistical techniques have just begun. The exclusive focus on geography as explanation for variation has given way to analyses combining geographical, linguistic, and social factors underlying language variation. Dialectometry has likewise ventured into diachronic studies and has also contributed theoretically to comparative dialectology and the study of dialect diffusion. Although the bulk of research involves lexis and phonology, morphosyntax is receiving increasing attention. Finally, new data sources and new (online) analytical software are expanding dialectometry's remit and its accessibility. 243

show abstract

A Cognitively Grounded Measure of Pronunciation Distance

Cited by 20 publications

References 20 publications

Opposing forces on acoustic duration

Opposing forces on acoustic duration

Modality effects in the cultural evolution of language: An experimental iterated learning approach

Advances in Dialectometry

Contact Info

Product

Resources

About