Search

Your search keyword '"Soskuthy, Marton"' showing total 20 results
Start Over Author "Soskuthy, Marton"
20 results on '"Soskuthy, Marton"'

1. Efficiently Identifying Low-Quality Language Subsets in Multilingual Datasets: A Case Study on a Large-Scale Multilingual Audio Dataset

Author

Samir, Farhan, Ahn, Emily P., Prakash, Shreya, Soskuthy, Márton, Shwartz, Vered, and Zhu, Jian

Subjects
Computer Science - Computation and Language
Abstract

Curating datasets that span multiple languages is challenging. To make the collection more scalable, researchers often incorporate one or more imperfect classifiers in the process, like language identification models. These models, however, are prone to failure, resulting in some language subsets being unreliable for downstream tasks. We introduce a statistical test, the Preference Proportion Test, for identifying such unreliable subsets. By annotating only 20 samples for a language subset, we're able to identify systematic transcription errors for 10 language subsets in a recent large multilingual transcribed audio dataset, X-IPAPack (Zhu et al., 2024). We find that filtering this low-quality data out when training models for the downstream task of phonetic transcription brings substantial benefits, most notably a 25.7% relative improvement on transcribing recordings in out-of-distribution languages. Our method lays a path forward for systematic and reliable multilingual dataset auditing., Comment: 16 pages, 6 figures

Published
2024

3. Generalised additive mixed models for dynamic analysis in linguistics: a practical introduction

Author

Sóskuthy, Márton

Subjects
Statistics - Applications
Abstract

This is a hands-on introduction to Generalised Additive Mixed Models (GAMMs) in the context of linguistics with a particular focus on dynamic speech analysis (e.g. formant contours, pitch tracks, diachronic change, etc.). The main goal is to explain some of the main ideas underlying GAMMs, and to provide a practical guide to frequentist significance testing using these models. The introduction covers a range of topics including basis functions, the smoothing penalty, random smooths, difference smooths, smooth interactions, model comparison and autocorrelation. It is divided into two parts. The first part looks at what GAMMs are, how they work and why/when we should use them. Although the reader can replicate some of the example analyses in this section, this is not essential. The second part is a tutorial introduction that illustrates the process of fitting and evaluating GAMMs in the R statistical software environment, and the reader is strongly encouraged to work through the examples on their own machine.

Published
2017

8. Phonetic biases and systemic effects in the actuation of sound change

Author

Soskuthy, Marton, Honeybone, Patrick, Kirby, Simon, and Kirby, James

Subjects
414, phonetic biases, phonetics, actuation problem, computational modelling
Abstract

This thesis investigates the role of phonetic biases and systemic effects in the actuation of sound change through computer simulations and experimental methods. Phonetic biases are physiological and psychoacoustic constraints on speech. One example is vowel undershoot: vowels sometimes fail to reach their phonetic targets due to limitations on the speed of the articulators. Phonetic biases are often paralleled by phonological patterns. For instance, many languages exhibit vowel reduction, a phonologised version of undershoot. To account for these parallels, a number of researchers have proposed that phonetic biases are the causal drive behind sound change. Although this proposal seems to solve the problem of actuation, its success is only apparent: while it might be able to explain situations where sound change occurs, it cannot easily explain the lack of sound change, that is, stasis. Since stability in sound systems seems to be the rule rather than the exception, the bias-based approach cannot provide an adequate account of their diachronic development on its own. The problem of bias-based accounts stems from their focus on changes affecting individual sound categories, and their neglect of system-wide interactions. The factors that affect speech production and perception define an adaptive landscape. The development of sound systems follows the topology of this landscape. When only a single category is investigated, it is easy to take an overly simplistic view of this landscape, and assume that phonetic biases are the only relevant factor. It is natural that the predicted outcomes will be simple and deterministic if such an approach is adopted. However, when we look at an entire sound system, other pressures such as contrast maintenance also become relevant, and the range of possible outcomes is much more diverse. Phonetic biases can still skew the adaptive landscape towards themselves, making phonetically natural outcomes more likely. However, their effects will often be countered by other pressures, which means that they will not be satisfied in every case. Sound systems move towards peaks in the adaptive landscape, or local optima, where the different pressures balance each other out. As a result, the system-based approach predicts stability. This stability can be broken by changes in the pressures that define the adaptive landscape. For instance, an increase or a decrease in functional load or a change in lexical distributions can create a situation where the sound system is knocked out of an equilibrium and starts evolving towards a new stable state. In essence, the adaptive landscape can create a moving target for the sound system. This ensures that both stability and change are observed. Therefore, this account makes realistic predictions with respect to the actuation problem. This argument is developed through a series of computer simulations that follow changes in artificial sound systems. All of these simulations are based on four theoretical assumptions: (i) speech production and perception are based on probabilistic category representations; (ii) these category representations are subject to continuous update throughout the lifetime of an individual; (iii) speech production and perception are affected by low-level universal phonetic biases; and (iv) category update is inhibited in cases where too many ambiguous tokens are produced due to category overlap. Special care is taken to anchor each of these assumptions in empirical results from a variety of fields including phonetics, sociolinguistics and psycholinguistics. Moreover, in order to show that the results described above follow directly from these theoretical assumptions and not other aspects of these models, the thesis demonstrates that exemplar and prototype models produce the same dynamics with respect to the observations above, and that the number of speakers in the model also does not have a significant influence on the outcomes. Much of the thesis focuses on rather abstract properties of simulated systems, which are difficult to test in a systematic way. The last chapter complements this by presenting a concrete example, which shows how the simulations can be linked to empirical data. Specifically, I look at the effect of lexical factors on the strength of contextual effects in sound categories, using the example of the voicing effect, whereby vowels are longer before voiced obstruents than they are before voiceless ones. The simulations implemented in this chapter predict a larger effect in cases where a given vowel category occurs equally frequently in voiced and voiceless environments, and a smaller difference where one of the environments dominates the lexical distribution of the vowel. This prediction is borne out in a small cross-linguistic production experiment looking at voicingconditioned vowel length patterns in French, Hungarian and English. Although this is only one of many predictions that fall out of the theory of sound change developed in this thesis, the success of this experiment is a strong indication that the research questions it brings into focus are worth investigating.

Published
2013

10. Multidimensional signals and analytic flexibility: Estimating degrees of freedom in human speech analyses

Author

Coretta, Stefano, Casillas, Joseph V, Roessig, Simon, Franke, Michael, Ahn, Byron, Al-Hoorie, Ali H., Al-Tamimi, Jalal, Alotaibi, Najd E., AlShakhori, Mohammed K., Altmiller, Ruth M., Arantes, Pablo, Athanasopoulou, Angeliki, Baese-Berk, Melissa M., Bailey, George, Baira A Sangma, Cheman, Beier, Eleonora J., Benavides, Gabriela M., Benker, Nicole, BensonMeyer, Emelia P., Benway, Nina R., Berry, Grant M., Bing, Liwen, Bjorndahl, Christina, Bolyanatz, Mariska, Braver, Aaron, Brown, Violet A., Brown, Alicia M., Brugos, Alejna, Buchanan, Erin M., Butlin, Tanna, Buxo-Lugo, Andres, Caillol, Coline, Cangemi, Francesco, Carignan, Christopher, Carraturo, Sita, Caudrelier, Tiphaine, Chodroff, Eleanor, Cohn, Michelle, Cronenberg, Johanna, Crouzet, Olivier, Dagar, Erica L., Dawson, Charlotte, Diantoro, Carissa A., Dokovova, Marie, Drake, Shiloh, Du, Fengting, Dubuis, Margaux, Dueme, Florent, Durward, Matthew, Egurtzegi, Ander, Elsherif, Mahmoud M., Esser, Janina, Ferragne, Emmanuel, Ferreira, Fernanda, Fink, Lauren K., Finley, Sara, Foster, Kurtis, Foulkes, Paul, Franzke, Rosa, Frazer-McKee, Gabriel, Fromont, Robert, Garcia, Christina, Geller, Jason, Grasso, Camille L., Greca, Pia, Grice, Martine, Grose-Hodge, Magdalena S., Gully, Amelia J., Halfacre, Caitlin, Hauser, Ivy, Hay, Jen, Haywood, Robert, Hellmuth, Sam, Hilger, Allison I., Holliday, Nicole, Hoogland, Damar, Huang, Yaqian, Hughes, Vincent, Icardo Isasa, Ane, Ilchovska, Zlatomira G., Jeon, Hae-Sung, Jones, Jacq, Junges, Magat N., Kaefer, Stephanie, Kaland, Constantijn, Kelley, Matthew C., Kelly, Niamh E., Kettig, Thomas, Khattab, Ghada, Koolen, Ruud, Krahmer, Emiel, Krajewska, Dorota, Krug, Andreas, Kumar, Abhilasha A., Lander, Anna, Lentz, Tomas O., Li, Wanyin, Li, Yanyu, Lialiou, Maria, Lima Jr., Ronaldo M., Lo, Justin J. H., Lopez Otero, Julio Cesar, Mackay, Bradley, MacLeod, Bethany, Mallard, Mel, McConnellogue, Carol-Ann Mary, Moroz, George, Murali, Mridhula, Nalborczyk, Ladislas, Nenadic, Filip, Nieder, Jessica, Nikolic, Dusan, Nogueira, Francisco G. S., Offerman, Heather M., Passoni, Elisa, Pelissier, Maud, Perry, Scott J., Pfiffner, Alexandra M., Proctor, Michael, Rhodes, Ryan, Rodriguez, Nicole, Roepke, Elizabeth, Roer, Jan P., Sbacco, Lucia, Scarborough, Rebecca, Schaeffler, Felix, Schleef, Erik, Schmitz, Dominic, Shiryaev, Alexander, Soskuthy, Marton, Spaniol, Malin, Stanley, Joseph, Strickler, Alyssa, Tavano, Alessandro, Tomaschek, Fabian, Tucker, Benjamin V., Turnbull, Rory, Ugwuanyi, Kingsley O., Urrestarazu-Porta, Inigo, van de Vijver, Ruben, Van Engen, Kristin J., van Miltenburg, Emiel, Wang, Bruce, Warner, Natasha, Wehrle, Simon, Westerbeek, Hans, Wiener, Seth, Winters, Stephen, Wong, Sidney G.-J., Wood, Anna, Wottawa, Jane, Xu, Chenzi, Zarate-Sanchez, German, Zellou, Georgia, Zhang, Cong, Zhu, Jian, Roettger, Timo B., Coretta, Stefano, Casillas, Joseph V, Roessig, Simon, Franke, Michael, Ahn, Byron, Al-Hoorie, Ali H., Al-Tamimi, Jalal, Alotaibi, Najd E., AlShakhori, Mohammed K., Altmiller, Ruth M., Arantes, Pablo, Athanasopoulou, Angeliki, Baese-Berk, Melissa M., Bailey, George, Baira A Sangma, Cheman, Beier, Eleonora J., Benavides, Gabriela M., Benker, Nicole, BensonMeyer, Emelia P., Benway, Nina R., Berry, Grant M., Bing, Liwen, Bjorndahl, Christina, Bolyanatz, Mariska, Braver, Aaron, Brown, Violet A., Brown, Alicia M., Brugos, Alejna, Buchanan, Erin M., Butlin, Tanna, Buxo-Lugo, Andres, Caillol, Coline, Cangemi, Francesco, Carignan, Christopher, Carraturo, Sita, Caudrelier, Tiphaine, Chodroff, Eleanor, Cohn, Michelle, Cronenberg, Johanna, Crouzet, Olivier, Dagar, Erica L., Dawson, Charlotte, Diantoro, Carissa A., Dokovova, Marie, Drake, Shiloh, Du, Fengting, Dubuis, Margaux, Dueme, Florent, Durward, Matthew, Egurtzegi, Ander, Elsherif, Mahmoud M., Esser, Janina, Ferragne, Emmanuel, Ferreira, Fernanda, Fink, Lauren K., Finley, Sara, Foster, Kurtis, Foulkes, Paul, Franzke, Rosa, Frazer-McKee, Gabriel, Fromont, Robert, Garcia, Christina, Geller, Jason, Grasso, Camille L., Greca, Pia, Grice, Martine, Grose-Hodge, Magdalena S., Gully, Amelia J., Halfacre, Caitlin, Hauser, Ivy, Hay, Jen, Haywood, Robert, Hellmuth, Sam, Hilger, Allison I., Holliday, Nicole, Hoogland, Damar, Huang, Yaqian, Hughes, Vincent, Icardo Isasa, Ane, Ilchovska, Zlatomira G., Jeon, Hae-Sung, Jones, Jacq, Junges, Magat N., Kaefer, Stephanie, Kaland, Constantijn, Kelley, Matthew C., Kelly, Niamh E., Kettig, Thomas, Khattab, Ghada, Koolen, Ruud, Krahmer, Emiel, Krajewska, Dorota, Krug, Andreas, Kumar, Abhilasha A., Lander, Anna, Lentz, Tomas O., Li, Wanyin, Li, Yanyu, Lialiou, Maria, Lima Jr., Ronaldo M., Lo, Justin J. H., Lopez Otero, Julio Cesar, Mackay, Bradley, MacLeod, Bethany, Mallard, Mel, McConnellogue, Carol-Ann Mary, Moroz, George, Murali, Mridhula, Nalborczyk, Ladislas, Nenadic, Filip, Nieder, Jessica, Nikolic, Dusan, Nogueira, Francisco G. S., Offerman, Heather M., Passoni, Elisa, Pelissier, Maud, Perry, Scott J., Pfiffner, Alexandra M., Proctor, Michael, Rhodes, Ryan, Rodriguez, Nicole, Roepke, Elizabeth, Roer, Jan P., Sbacco, Lucia, Scarborough, Rebecca, Schaeffler, Felix, Schleef, Erik, Schmitz, Dominic, Shiryaev, Alexander, Soskuthy, Marton, Spaniol, Malin, Stanley, Joseph, Strickler, Alyssa, Tavano, Alessandro, Tomaschek, Fabian, Tucker, Benjamin V., Turnbull, Rory, Ugwuanyi, Kingsley O., Urrestarazu-Porta, Inigo, van de Vijver, Ruben, Van Engen, Kristin J., van Miltenburg, Emiel, Wang, Bruce, Warner, Natasha, Wehrle, Simon, Westerbeek, Hans, Wiener, Seth, Winters, Stephen, Wong, Sidney G.-J., Wood, Anna, Wottawa, Jane, Xu, Chenzi, Zarate-Sanchez, German, Zellou, Georgia, Zhang, Cong, Zhu, Jian, and Roettger, Timo B.

Abstract

Recent empirical studies have highlighted the large degree of analytic flexibility in data analysis which can lead to substantially different conclusions based on the same data set. Thus, researchers have expressed their concerns that these researcher degrees of freedom might facilitate bias and can lead to claims that do not stand the test of time. Even greater flexibility is to be expected in fields in which the primary data lend themselves to a variety of possible operationalizations. The multidimensional, temporally extended nature of speech constitutes an ideal testing ground for assessing the variability in analytic approaches, which derives not only from aspects of statistical modeling, but also from decisions regarding the quantification of the measured behavior. In the present study, we gave the same speech production data set to 46 teams of researchers and asked them to answer the same research question, resulting insubstantial variability in reported effect sizes and their interpretation. Using Bayesian meta-analytic tools, we further find little to no evidence that the observed variability can be explained by analysts’ prior beliefs, expertise or the perceived quality of their analyses. In light of this idiosyncratic variability, we recommend that researchers more transparently share details of their analysis, strengthen the link between theoretical construct and quantitative system and calibrate their (un)certainty in their conclusions.

Published
2023

18. When the tune shapes morphology: the origins of vocatives

Author

Soskuthy , Marton and Roettger , Timo

Subjects
PsyArXiv|Social and Behavioral Sciences, bepress|Social and Behavioral Sciences, bepress|Social and Behavioral Sciences|Linguistics, PsyArXiv|Social and Behavioral Sciences|Linguistics
Abstract

Many languages use pitch to express pragmatic meaning (henceforth “tune”). This requires segmental carriers with rich harmonic structure and high periodic energy, making vowels the optimal carriers of the tune. Tunes can be phonetically impoverished when there is a shortage of vowels, endangering the recovery of their function. This biases sound systems towards the optimisation of tune transmission by processes such as the insertion of vowels. Vocative constructions – used to attract and maintain the addressee’s attention – are often characterised by specific tunes. Many languages additionally mark vocatives morphologically. In this paper, we argue that one potential pathway for the emergence of vocative morphemes is the grammaticalisation of tune-driven phonetic variation that helps to carry pitch patterns. Looking at a corpus of 101 languages, we compare vocatives to structural case markers in terms of their phonological make-up. We find that vocatives are often characterised by additional prosodic modulation (vowel lengthening, stress shift, tone change) and contain substantially fewer consonants, supporting our hypothesis that the acoustic properties of tunes interact with segmental features and can shape the emergence of morphological markers. This fits with the view that the efficient transmission of information is a driving force in the evolution of languages, but also highlights the importance of defining “information” broadly to include pragmatic, social and affectual components alongside propositional meaning.

Published
2019

20. A dynamic acoustic view of real-time change in word-final liquids in spontaneous Glaswegian

Author

Stuart-Smith, Jane, Lennon, Robert, Macdonald, Rachel, Robertson, Duncan, Soskuthy, Marton, Jose, Brian, and Evers, Ludger

Subjects
PE, P1
Abstract

This paper investigates the acoustic evidence for real-time change in word-final liquids (/r/ and /l/) in a small-scale study of older male Glaswegian speakers recorded from the 1970s to the 2000s. A dynamic acoustic analysis of the first three formants across the duration of the rhyme (vowel+liquid sequence) shows significant effects of preceding and \ud following phonetic context on the course and trajectories of the formant tracks. We also find raising of F3 for /r/ in speakers who were born and recorded more recently; F2 is lowering for /l/ in the same speakers. Comparison of F2 across the two word-final liquids suggests that /r/ is clearer than /l/ for this Scottish dialect; interestingly the polarity in resonance between /r/ and /l/ is increasing over time.

Catalog

Books, media, physical & digital resources
See catalog results