Understanding the Phonetic Characteristics of Speech Under Uncertainty—Implications of the Representation of Linguistic Knowledge in Learning and Processing

Abstract

The uncertainty associated with paradigmatic families has been shown to correlate with their phonetic characteristics in speech, suggesting that representations of complex sublexical relations between words are part of speaker knowledge. To better understand this, recent studies have used two-layer neural network models to examine the way paradigmatic uncertainty emerges in learning. However, to date this work has largely ignored the way choices about the representation of inflectional and grammatical functions (IFS) in models strongly influence what they subsequently learn. To explore the consequences of this, we investigate how representations of IFS in the input-output structures of learning models affect the capacity of uncertainty estimates derived from them to account for phonetic variability in speech. Specifically, we examine whether IFS are best represented as outputs to neural networks (as in previous studies) or as inputs by building models that embody both choices and examining their capacity to account for uncertainty effects in the formant trajectories of word final [ɐ], which in German discriminates around sixty different IFS. Overall, we find that formants are enhanced as the uncertainty associated with IFS decreases. This result dovetails with a growing number of studies of morphological and inflectional families that have shown that enhancement is associated with lower uncertainty in context. Importantly, we also find that in models where IFS serve as inputs—as our theoretical analysis suggests they ought to—its uncertainty measures provide better fits to the empirical variance observed in [ɐ] formants than models where IFS serve as outputs. This supports our suggestion that IFS serve as cognitive cues during speech production, and should be treated as such in modeling. It is also consistent with the idea that when IFS serve as inputs to a learning network. This maintains the distinction between those parts of the network that represent message and those that represent signal. We conclude by describing how maintaining a “signal-message-uncertainty distinction” can allow us to reconcile a range of apparently contradictory findings about the relationship between articulation and uncertainty in context.

Modelling German Word Stress

“…The functionality of the two algorithms has been successfully demonstrated to capture discriminative learning (Bröker & Ramscar 2020;Ramscar 2021) of morphological processes including inflection (Ramscar & Yarlett 2007;Ramscar et al 2013bRamscar et al , 2010Nieder et al 2021Nieder et al , 2022, of processing in the context of reading and listening to morphological simple and complex words (Baayen et al 2011;Arnold et al 2017), and also in other domains such as the learning of phonetic categories (Olejarczuk et al 2018;Nixon 2020;Nixon & Tomaschek 2020 and speech production (Ramscar & Yarlett 2007;Tomaschek et al 2019;Baayen et al 2019;Tomaschek & Ramscar 2022). An introduction to using NDL can be found in Tomaschek (2020) and an excellent overview of how the dynamics of connection weights depend on cue-to-outcome constellations is presented by Hoppe et al (2022).…”

The keys to the future? An examination of associative versus discriminative accounts of Serial Pattern Learning

“…Discriminative learning process thus capture incremental, implicit learning, rather than explicit changes in perception or beliefs based on logic or reasoning (see e.g., Ramscar et al, 2013a;Nixon, Poelstra and van Rij, 2022), and can be formalised by training neural networks using virtually any computational learning algorithm that takes into account error during learning (though whether a model is discriminative or associative depends on the representation of cues and outcomes in the model, see Bröker and Ramscar, 2020). Because they are relatively simple to implement, and models based on them are easily interpreted, a number of recent psycholinguistic studies have made use of the learning equations proposed by Rescorla and Wagner (1972) to formalise learning problems and derive predictions about participants' behaviour during experiments in domains such as: word learning (Nixon, 2020;Ramscar et al, 2010;Ramscar, Dye, Popick and O'Donnell-McCarthy, 2011;Ramscar et al, 2013a), morphological structure learning in children (Ramscar et al, 2010(Ramscar et al, , 2011(Ramscar et al, , 2013b and adults (Arnon and Ramscar, 2012;Ramscar, 2013;Vujović, Ramscar and Wonnacott, 2021); morphological processing in adults (Baayen, Milin, Durdevic, Hendrix and Marelli, 2011;Tomaschek, Plag, Ernestus and Baayen, 2019;Nieder, Tomaschek, Cohrs and de Vijver, 2021); auditory comprehension and recognition (Arnold, Tomaschek, Sering, Lopez and Baayen, 2017;Shafaei-Bajestan and Baayen, 2018); effects of the lexicon on articulation (Tucker, Sims and Baayen, 2019;Tomaschek et al, 2019;Schmitz, Plag, Baer-Henney and Stein, 2021;Stein and Plag, 2021;Tomaschek and Ramscar, 2022); phonetic learning (Nixon, 2018(Nixon, , 2020Tomaschek, 2020, 2021); and the trial-by-trial changes in the neural signal that occur with learning…”

“…In first language acquisition, Tomaschek (2020, 2021) present a computational model of early infants' learning of speech by using the incoming acoustic signal to predict upcoming acoustic signal. Apart from phonetic learning, error-driven learning has also been found to play a role in word learning (Ramscar et al, 2013a(Ramscar et al, , 2010(Ramscar et al, , 2011, morphological learning (Ramscar and Yarlett, 2007;Ramscar et al, 2013b;Tomaschek et al, 2019;Hoppe et al, 2020) and speech production (Tucker et al, 2019;Tomaschek and Ramscar, 2022) and speech perception (Shafaei-Bajestan and .There is also evidence for trial-by-trial error-driven learning in the brain (Lentz et al, 2022) and during lexical decision (? ).…”

Stratification effects without morphological strata, syllable counting effects without counts – modelling English stress assignment with Naive Discriminative Learning

“…In addition, the error-driven learning rule successfully models aspects of child language acquisition (Ramscar et al, 2010(Ramscar et al, , 2011(Ramscar et al, , 2013b, acquisition and usage of allomorphic suffixes (Divjak et al, 2021), reaction times in lexical decision tasks (Baayen et al, 2011;Milin et al, 2017b), self-paced reading (Milin et al, 2017a), phonetic characteristics depending on their morphological function (Tomaschek et al, 2019;Saito et al, 2020;Tomaschek & Ramscar, 2022;Schmitz et al, 2021), auditory comprehension (Baayen et al, 2016;Arnold et al, 2017) and acoustic single-word recognition (Shafaei-Bajestan & Baayen, 2018). Furthermore, it was applied in modelling early phonetic learning (Nixon & Tomaschek, 2021 and morphological processes of pluralization in Maltese (Nieder et al, 2022a,b).…”