Publications

Abstract

Word count estimation (WCE) from audio recordings has a number of applications, including quantifying the amount of speech that language-learning infants hear in their natural environments, as captured by daylong recordings made with devices worn by infants. To be applicable in a wide range of scenarios and also low-resource domains, WCE tools should be extremely robust against varying signal conditions and require minimal access to labeled training data in the target domain. For this purpose, earlier work has used automatic syllabification of speech, followed by a least-squares-mapping of syllables to word counts. This paper compares a number of previously proposed syllabifiers in the WCE task, including a supervised bi-directional long short-term memory (BLSTM) network that is trained on a language for which high quality syllable annotations are available (a “high resource language”), and reports how the alternative methods compare on different languages and signal conditions. We also explore additive noise and varying-channel data augmentation strategies for BLSTM training, and show how they improve performance in both matching and mismatching languages. Intriguingly, we also find that even though the BLSTM works on languages beyond its training data, the unsupervised algorithms can still outperform it in challenging signal conditions on novel languages.

Abstract

We suggest that it is now possible to conduct research on a topic which might be called evolutionary geophonetics. The main question is how the climate influences the evolution of language. This involves biological adaptations to the climate that may affect biases in production and perception; cultural evolutionary adaptations of the sounds of a language to climatic conditions; and influences of the climate on language diversity and contact. We discuss these ideas with special reference to a recent hypothesis that lexical tone is not adaptive in dry climates (Everett, Blasi & Roberts, 2015).

Abstract

For almost two decades, the poor performance observed with the so-called Director task has been interpreted as evidence of limited use of Theory of Mind in communication. Here we propose a probabilistic model of common ground in referential communication that derives three inferences from an utterance: what the speaker is talking about in a visual context, what she knows about the context, and what referential expressions she prefers. We tested our model by comparing its inferences with those made by human participants and found that it closely mirrors their judgments, whereas an alternative model compromising the hearer’s expectations of cooperativeness and efficiency reveals a worse fit to the human data. Rather than assuming that common ground is fixed in a given exchange and may or may not constrain reference resolution, we show how common ground can be inferred as part of the process of reference assignment.

Abstract

While there are many studies on information retrieval models using full-text, there are presently no comparison studies of full-text retrieval vs. retrieval only over the titles of documents. On the one hand, the full-text of documents like scientific papers is not always available due to, e.g., copyright policies of academic publishers. On the other hand, conducting a search based on titles alone has strong limitations. Titles are short and therefore may not contain enough information to yield satisfactory search results. In this paper, we compare different retrieval models regarding their search performance on the full-text vs. only titles of documents. We use different datasets, including the three digital library datasets: EconBiz, IREON, and PubMed. The results show that it is possible to build effective title-based retrieval models that provide competitive results comparable to full-text retrieval. The difference between the average evaluation results of the best title-based retrieval models is only 3% less than those of the best full-text-based retrieval models.

Abstract

A series of experiment tested Ekman’s (1992) hypothesis that there are a set of positive basic emotions that are expressed using vocal para-linguistic sounds, e.g. laughter and cheers. The proposed categories investigated were amusement, contentment, pleasure, relief and triumph. Behavioural testing using a forced-choice task indicated that participants were able to reliably recognize vocal expressions of the proposed emotions. A cross-cultural study in the preliterate Himba culture in Namibia confirmed that these categories are also recognized across cultures. A recognition test of acoustically manipulated emotional vocalizations established that the recognition of different emotions utilizes different vocal cues, and that these in turn differ from the cues used when comprehending speech. In a study using fMRI we found that relative to a signal correlated noise baseline, the paralinguistic expressions of emotion activated bilateral superior temporal gyri and sulci, lateral and anterior to primary auditory cortex, which is consistent with the processing of non linguistic vocal cues in the auditory ‘what’ pathway. Notably amusement was associated with greater activation extending into both temporal poles and amygdale and insular cortex. Overall, these results support the claim that ‘happiness’ can be fractionated into amusement, pleasure, relief and triumph.

Abstract

Fine-Tracker is a speech-based model of human speech
recognition. While previous work has shown that Fine-Tracker
is successful at modelling aspects of human spoken-word
recognition, its speech recognition performance is not
comparable to that of human performance, possibly due to
suboptimal intermediate articulatory feature (AF)
representations. This study investigates the effect of improved
AF representations, obtained using a state-of-the-art deep
convolutional network, on Fine-Tracker’s simulation and
recognition performance: Although the improved AF quality
resulted in improved speech recognition; it, surprisingly, did
not lead to an improvement in Fine-Tracker’s simulation power.

Abstract

This paper addresses the issue of large-vocabulary recognition in a specific word class. We propose a two-pass strategy in which only major cities are explicitly represented in the first stage lexicon. An unknown word model encoded as a phone loop is used to detect OOV city names (referred to as rare city names). After which SpeM, a tool that can extract words and word-initial cohorts from phone graphs on the basis of a large fallback lexicon, provides an N-best list of promising city names on the basis of the phone sequences generated in the first stage. This N-best list is then inserted into the second stage lexicon for a subsequent recognition pass. Experiments were conducted on a set of spontaneous telephone-quality utterances each containing one rare city name. We tested the size of the N-best list and three types of language models (LMs). The experiments showed that SpeM was able to include nearly 85% of the correct city names into an N-best list of 3000 city names when a unigram LM, which also boosted the unigram scores of a city name in a given state, was used.

Abstract

In this paper, we illustrate the close parallels between the research fields of human speech recognition (HSR) and automatic speech recognition (ASR) using a computational model of human word recognition, SpeM, which was built using techniques from ASR. We show that ASR has proven to be useful for improving models of HSR by relieving them of some of their shortcomings. However, in order to build an integrated computational model of all aspects of HSR, a lot of issues remain to be resolved. In this process, ASR algorithms and techniques definitely can play an important role.

Abstract

Processing of semantic information in language comprehension has been suggested to be modulated by attentional resources. Consequently, cognitive load would be expected to reduce semantic priming, but studies have yielded inconsistent results. This study investigated whether cognitive load affects semantic activation in speech processing in older adults, and whether this is modulated by individual differences in cognitive and hearing abilities. Older adults participated in an auditory continuous lexical decision task in a low-load and high-load condition. The group analysis showed only a marginally significant reduction of semantic priming in the high-load condition compared to the low-load condition. The individual differences analysis showed that semantic priming was significantly reduced under increased load in participants with poorer attention-switching control. Hence, a resource-demanding secondary task may affect the integration of spoken words into a coherent semantic representation for listeners with poorer attentional skills.

Abstract

Speakers can adapt their speech and co-speech gestures for
addressees. Here, we investigate whether this ability is
modulated by age. Younger and older adults participated in a
comic narration task in which one participant (the speaker)
narrated six short comic stories to another participant (the
addressee). One half of each story was known to both participants, the other half only to the speaker. Younger but
not older speakers used more words and gestures when narrating novel story content as opposed to known content.
We discuss cognitive and pragmatic explanations of these findings and relate them to theories of gesture production.

Abstract

Adaptation to altered auditory feedback has been shown to induce subsequent shifts in perception. However, it is uncertain whether these perceptual changes may generalize to other speech sounds. In this experiment, we tested whether exposing the production of a vowel to altered auditory feedback affects perceptual categorization of a consonant distinction. In two sessions, participants produced CVC words containing the vowel /i/, while intermittently categorizing stimuli drawn from a continuum between "see" and "she." In the first session feedback was unaltered, while in the second session the formants of the vowel were shifted 20% towards /u/. Adaptation to the altered vowel was found to reduce the proportion of perceived /S/ stimuli. We suggest that this reflects an alteration to the sensorimotor mapping that is shared between vowels and consonants.

Abstract

Speakers adapt their speech and gestures in various ways for their audience. We investigated further whether they use
ostensive signals (eye gaze, ostensive speech (e.g. like this, this) or a combination of both) in relation to their gestures
when talking to different addressees, i.e., to another adult or a child in a multimodal demonstration task. While adults used
more eye gaze towards their gestures with other adults than with children, they were more likely to use combined
ostensive signals for children than for adults. Thus speakers mark the communicative relevance of their gestures with different types of ostensive signals and by taking different types of addressees into account.

Abstract

Previous research [9] suggests that Dutch learners of (British) English are not able to express sarcasm prosodically in their L2. The present study investigates whether explicit training on the prosodic markers of sarcasm in English can improve learners’ realisation of sarcasm. Sarcastic speech was elicited in short simulated telephone conversations between Dutch advanced learners of English and a native British English-speaking ‘friend’ in two sessions, fourteen days apart. Between the two sessions, participants were trained by means of (1) a presentation, (2) directed independent practice, and (3) evaluation of participants’ production and individual feedback in small groups. L1 British English-speaking raters subsequently evaluated the degree of sarcastic sounding in the participants’ responses on a five-point scale. It was found that significantly higher sarcasm ratings were given to L2 learners’ production obtained after the training than that obtained before the training; explicit training on prosody has a positive effect on learners’ production of sarcasm.

Abstract

We report an individual with music-odor synaesthesia who experiences automatic and vivid odor sensations when she hears music. S’s odor associations were recorded on two days, and compared with those of two control participants. Overall, S produced longer descriptions, and her associations were of multiple odors at once, in comparison to controls who typically reported a single odor. Although odor associations were qualitatively different between S and controls, ratings of the consistency of their descriptions did not differ. This demonstrates that crossmodal associations between music and odor exist in non-synaesthetes too. We also found that S is better at discriminating between odors than control participants, and is more likely to experience emotion, memories and evaluations triggered by odors, demonstrating the broader impact of her synaesthesia.

Abstract

Sequences of reaction times (RT) produced by participants in an experiment are not only influenced by the stimuli, but by many other factors as well, including fatigue, attention, experience, IQ, handedness, etc. These confounding factors result in longterm effects (such as a participant’s overall reaction capability) and in short- and medium-time fluctuations in RTs (often referred to as ‘local speed effects’). Because stimuli are usually presented in a random sequence different for each participant, local speed effects affect the underlying ‘true’ RTs of specific trials in different ways across participants. To be able to focus statistical analysis on the effects of the cognitive process under study, it is necessary to reduce the effect of confounding factors as much as possible. In this paper we propose and compare techniques and criteria for doing so, with focus on reducing (‘filtering’) the local speed effects. We show that filtering matters substantially for the significance analyses of predictors in linear mixed effect regression models. The performance of filtering is assessed by the average between-participant correlation between filtered RT sequences and by Akaike’s Information Criterion, an important measure of the goodness-of-fit of linear mixed effect regression models.

Abstract

This paper presents DIANA, a new computational model of human speech processing. It is the first model that simulates the complete processing chain from the on-line processing of an acoustic signal to the execution of a response, including reaction times. Moreover it assumes minimal modularity. DIANA consists of three components. The activation component computes a probabilistic match between the input acoustic signal and representations in DIANA’s lexicon, resulting in a list of word hypotheses changing over time as the input unfolds. The decision component operates on this list and selects a word as soon as sufficient evidence is available. Finally, the execution component accounts for the time to execute a behavioral action. We show that DIANA well simulates the average participant in a word recognition experiment.

Abstract

DIANA is an end-to-end computational model of speech processing, which takes as input the speech signal, and provides as output the orthographic transcription of the stimulus, a word/non-word judgment and the associated estimated reaction time. So far, the model has only been tested for Dutch. In this paper, we extend DIANA such that it can also process North American English. The model is tested by having it simulate human participants in a large scale North American English lexical decision experiment. The simulations show that DIANA can adequately approximate the reaction times of an average participant (r = 0.45). In addition, they indicate that DIANA does not yet adequately model the cognitive processes that take place after stimulus offset.

Abstract

This paper investigates the interaction between acoustic scores and symbolic mismatch penalties in multi-pass speech decoding techniques that are based on the creation of a segment graph followed by a lexical search. The interaction between acoustic and symbolic mismatches determines to a large extent the structure of the search space of these multipass approaches. The background of this study is a recently developed computational model of human word recognition, called SpeM. SpeM is able to simulate human word recognition data and is built as a multi-pass speech decoder. Here, we focus on unravelling the structure of the search space that is used in SpeM and similar decoding strategies. Finally, we elaborate on the close relation between distances in this search space, and distance measures in search spaces that are based on a combination of acoustic and phonetic features.

Abstract

The recent advent of deep learning techniques in speech tech-nology and in particular in automatic speech recognition hasyielded substantial performance improvements. This suggeststhat deep neural networks (DNNs) are able to capture structurein speech data that older methods for acoustic modeling, suchas Gaussian Mixture Models and shallow neural networks failto uncover. In image recognition it is possible to link repre-sentations on the first couple of layers in DNNs to structuralproperties of images, and to representations on early layers inthe visual cortex. This raises the question whether it is possi-ble to accomplish a similar feat with representations on DNNlayers when processing speech input. In this paper we presentthree different experiments in which we attempt to untanglehow DNNs encode speech signals, and to relate these repre-sentations to phonetic knowledge, with the aim to advance con-ventional phonetic concepts and to choose the topology of aDNNs more efficiently. Two experiments investigate represen-tations formed by auto-encoders. A third experiment investi-gates representations on convolutional layers that treat speechspectrograms as if they were images. The results lay the basisfor future experiments with recursive networks.

Abstract

Apraxia of Speech (AOS) is a motor speech disorder whose precise nature is still poorly understood. A recent behavioural experiment featuring a noise masking paradigm suggests that AOS reflects a disruption of feedforward control, whereas feedback control is spared and plays a more prominent role in achieving and maintaining segmental contrasts [10]. In the present study, we set out to validate the interpretation of AOS as a feedforward impairment by means of a series of computational simulations with the DIVA model [6, 7] mimicking the behavioural experiment. Simulation results showed a larger reduction in vowel spacing and a smaller vowel dispersion in the masking condition compared to the no-masking condition for the simulated feedforward deficit, whereas the other groups showed an opposite pattern. These results mimic the patterns observed in the human data, corroborating the notion that AOS can be conceptualized as a deficit in feedforward control

Abstract

We estimate lexical Concreteness for millions of words across 77 languages. Using a simple regression framework, we combine vector-based models of lexical semantics with experimental norms of Concreteness in English and Dutch. By applying techniques to align vector-based semantics across distinct languages, we compute and release Concreteness estimates at scale in numerous languages for which experimental norms are not currently available. This paper lays out the technique and its efficacy. Although this is a difficult dataset to evaluate immediately, Concreteness estimates computed from English correlate with Dutch experimental norms at $\rho$ = .75 in the vocabulary at large, increasing to $\rho$ = .8 among Nouns. Our predictions also recapitulate attested relationships with word frequency. The approach we describe can be readily applied to numerous lexical measures beyond Concreteness

Abstract

Do all languages convey semantic knowledge in the same way? If language simply mirrors the structure of the world, the answer should be a qualified “yes”. If, however, languages impose structure as much as reflecting it, then even ostensibly the “same” word in different languages may mean quite different things. We provide a first pass at a large-scale quantification of cross-linguistic semantic alignment of approximately 1000 meanings in 55 languages. We find that the translation equivalents in some domains (e.g., Time, Quantity, and Kinship) exhibit high alignment across languages while the structure of other domains (e.g., Politics, Food, Emotions, and Animals) exhibits substantial cross-linguistic variability. Our measure of semantic alignment correlates with known phylogenetic distances between languages: more phylogenetically distant languages have less semantic alignment. We also find semantic alignment to correlate with cultural distances between societies speaking the languages, suggesting a rich co-adaptation of language and culture even in domains of experience that appear most constrained by the natural world

Abstract

This article describes how the tonal elements of two common Spanish intonation contours –the falling statement and the low-rising-falling request– align with the segmental string in broad-focus utterances differing in number of prosodic words. Using an imitation-and-completion task, we show that (i) the last stressed syllable of the utterance, traditionally viewed as carrying the ‘nuclear’ accent, associates with either a high or a low tonal element depending on phrase length (ii) that certain tonal elements can be realized or omitted depending on the availability of specific metrical positions in their intonational phrase, and (iii) that the high tonal element of the request contour associates with either a stressed syllable or an intonational phrase edge depending on phrase length. On the basis of these facts, and in contrast to previous descriptions of Spanish intonation relying on obligatory and constant nuclear contours (e.g., L* L% for all neutral statements), we argue for a less constrained intonational morphology involving tonal units linked to the segmental string via contour-specific principles.

Abstract

Violations of the maxims of Quantity occur when utterances provide more (over-specified) or less (under-specified) information than strictly required for referent identification. While behavioural datasuggest that under-specified expressions lead to comprehension difficulty and communicative failure, there is no consensus as to whether over-specified expressions are also detrimental to comprehension. In this study we shed light on this debate, providing neurophysiological evidence supporting the view that extra information facilitates comprehension. We further present novel evidence that referential failure due to under-specification is qualitatively different from explicit cases of referential failure, when no matching referential candidate is available in the context.

Abstract

In situated communication, reference to an entity in the shared visual context can be established using eitheranexpression that conveys precise (minimally specified) or redundant (over-specified) information. There is, however, along-lasting debate in psycholinguistics concerningwhether the latter hinders referential processing. We present evidence from an eyetrackingexperiment recordingfixations as well asthe Index of Cognitive Activity –a novel measure of cognitive workload –supporting the view that over-specifications facilitate processing. We further present originalevidence that, above and beyond the effect of specificity,referring expressions thatuniformly reduce referential entropyalso benefitprocessing

Abstract

The task of automatic playlist continuation is generating a list of recommended tracks that can be added to an existing playlist. By suggesting appropriate tracks, i. e., songs to add to a playlist, a recommender system can increase the user engagement by making playlist creation easier, as well as extending listening beyond the end of current playlist. The ACM Recommender Systems Challenge 2018 focuses on such task. Spotify released a dataset of playlists, which includes a large number of playlists and associated track listings. Given a set of playlists from which a number of tracks have been withheld, the goal is predicting the missing tracks in those playlists. We participated in the challenge as the team Unconscious Bias and, in this paper, we present our approach. We extend adversarial autoencoders to the problem of automatic playlist continuation. We show how multiple input modalities, such as the playlist titles as well as track titles, artists and albums, can be incorporated in the playlist continuation task.

Abstract

Languages combine arbitrary and iconic signals. How do iconic signals emerge and when do they persist? We present an experimental study of the role of iconicity in the emergence of structure in an artificial language. Using an iterated communication game in which we control the signalling medium as well as the meaning space, we study the evolution of communicative signals in transmission chains. This sheds light on how affordances of the communication medium shape and constrain the mappability and transmissibility of form-meaning pairs. We find that iconic signals can form the building blocks for wider compositional patterns

Abstract

Before infants become mature speakers of their native language, they must acquire a robust word-recognition system which allows them to strike the balance between allowing some variation (mood, voice, accent) and recognizing variability that potentially changes meaning (e.g. cat vs hat). The current meta-analysis quantifies how the latter, termed mispronunciation sensitivity, changes over infants’ first three years, testing competing predictions of mainstream language acquisition theories. Our results show that infants were sensitive to mispronunciations, but accepted them as labels for target objects. Interestingly, and in contrast to predictions of mainstream theories, mispronunciation sensitivity was not modulated by infant age, suggesting that a sufficiently flexible understanding of native language phonology is in place at a young age.

Abstract

Distributional vowel training for adults has been reported as “effective” for Spanish and Bulgarian learners of Dutch vowels, in studies using a behavioural task. A recent study did not yield a similar clear learning effect for Dutch learners of the English vowel contrast /æ/~/ε/, as measured with event-related potentials (ERPs). The present study aimed to examine the possibility that the latter result was related to the method. As in the ERP study, we tested whether distributional training improved Dutch adult learners’ perception of English /æ/~/ε/. However, we measured behaviour instead of ERPs, in a design identical to that used in the previous studies with Spanish learners. The results do not support an effect of distributional training and thus “replicate” the ERP study. We conclude that it remains unclear whether distributional vowel training is effective for Dutch adults.

Abstract

Recent studies using phoneme detection tasks have shown that spoken-language processing is neither facilitated nor interfered with by optional assimilation, but is inhibited by violation of obligatory assimilation. Interpretation of these results depends on an assessment of their generality, specifically, whether they also obtain when listeners are processing nonnative language. Two separate experiments are presented in which native listeners of German and native listeners of Dutch had to detect a target fricative in legal monosyllabic Dutch nonwords. All of the nonwords were correct realisations in standard Dutch. For German listeners, however, half of the nonwords contained phoneme strings which violate the German fricative assimilation rule. Whereas the Dutch listeners showed no significant effects, German listeners detected the target fricative faster when the German fricative assimilation was violated than when no violation occurred. The results might suggest that violation of assimilation rules does not have to make processing more difficult per se.

Abstract

In the word-learning domain, both adults and young children are able to find the correct referent of a word from highly ambiguous contexts that involve many words and objects by computing distributional statistics across the co-occurrences of words and referents at multiple naming moments (Yu & Smith, 2007; Smith & Yu, 2008). However, there is still debate regarding how learners accumulate distributional information to learn object labels in natural learning environments, and what underlying learning mechanism learners are most likely to adopt. Using the Human Simulation Paradigm (Gillette, Gleitman, Gleitman & Lederer, 1999), we found that participants’ learning performance gradually improved and that their ability to remember and carry over partial knowledge from past learning instances facilitated subsequent learning. These results support the statistical learning model that word learning is a continuous process.