Publications

Abstract

In this work we perform multimodal detection and classification
of head movements from face to face video conversation data.
We have experimented with different models and feature sets
and provided some insight on the effect of independent features,
but also how their interaction can enhance a head movement
classifier. Used features include nose, neck and mid hip position
coordinates and their derivatives together with acoustic features,
namely, intensity and pitch of the speaker on focus. Results
show that when input features are sufficiently processed by in-
teracting with each other, a linear classifier can reach a similar
performance to a more complex non-linear neural model with
several hidden layers. Our best models achieve state-of-the-art
performance in the detection task, measured by macro-averaged
F1 score.

Abstract

Continuous vector word representations (or
word embeddings) have shown success in cap-turing semantic relations between words, as evidenced by evaluation against behavioral data of adult performance on semantic tasks (Pereira et al., 2016). Adult semantic knowl-edge is the endpoint of a language acquisition process; thus, a relevant question is whether these models can also capture emerging word
representations of young language learners. However, the data for children’s semantic knowledge across development is scarce. In this paper, we propose to bridge this gap by using Age of Acquisition norms to evaluate word embeddings learnt from child-directed input. We present two methods that evaluate word embeddings in terms of (a) the semantic neighbourhood density of learnt words, and (b) con-
vergence to adult word associations. We apply our methods to bag-of-words models, and find that (1) children acquire words with fewer semantic neighbours earlier, and (2) young learners only attend to very local context. These findings provide converging evidence for validity of our methods in understanding the prerequisite features for a distributional model of word learning.

Abstract

We present the Retention and Recognition model (R&R), a probabilistic exemplar model that accounts for segmentation in Artificial Language Learning experiments. We show that R&R provides an excellent fit to human responses in three segmentation experiments with adults (Frank et al., 2010), outperforming existing models. Additionally, we analyze the results of the simulations and propose alternative explanations for the experimental findings.

Abstract

The present paper assesses discourse-pragmatic factors as a potential explanation for the subject-object assymetry in early child language. It identifies a set of factors which characterize typical situations of informativeness (Greenfield & Smith, 1976), and uses these factors to identify informative arguments in data from four children aged 2;0 through 3;6 learning Inuktitut as a first language. In addition, it assesses the extent of the links between features of informativeness on one hand and lexical vs. null and subject vs. object arguments on the other. Results suggest that a pragmatics account of the subject-object asymmetry can be upheld to a greater extent than previous research indicates, and that several of the factors characterizing informativeness are good indicators of those arguments which tend to be omitted in early child language.

Abstract

The term “uptalk” describes utterance-final pitch rises that carry no sentence-structural information. Uptalk is usually dialectal or sociolectal, and Australian English (AusEng) is particularly known for this attribute. We ask here whether experience with an uptalk variety affects listeners’ ability to categorise rising pitch contours on the basis of the timing and height of their onset and offset. Listeners were two groups of English-speakers (AusEng, and American English), and three groups of listeners with L2 English: one group with Mandarin as L1 and experience of listening to AusEng, one with German as L1 and experience of listening to AusEng, and one with German as L1 but no AusEng experience. They heard nouns (e.g. flower, piano) in the framework “Got a NOUN”, each ending with a pitch rise artificially manipulated on three contrasts: low vs. high rise onset, low vs. high rise offset and early vs. late rise onset. Their task was to categorise the tokens as “question” or “statement”, and we analysed the effect of the pitch contrasts on their judgements. Only the native AusEng listeners were able to use the pitch contrasts systematically in making these categorisations.

Abstract

The use of subject pronouns by bilingual speakers using both a pro-drop and a non-pro-drop language (e.g. Spanish heritage speakers in the USA) is a well-studied topic in research on cross-linguistic influence in language contact situations. Previous studies looking at bilinguals with different proficiency levels have yielded conflicting results on whether there is transfer from the non-pro-drop patterns to the pro-drop language. Additionally, previous research has focused on speech patterns only. In this paper, we study the two modalities of language, speech and gesture, and ask whether and how they reveal cross-linguistic influence on the use of subject pronouns in discourse. We focus on elicited narratives from heritage speakers of Turkish in the Netherlands, in both Turkish (pro-drop) and Dutch (non-pro-drop), as well as from monolingual control groups. The use of pronouns was not very common in monolingual Turkish narratives and was constrained by the pragmatic contexts, unlike in Dutch. Furthermore, Turkish pronouns were more likely to be accompanied by localized gestures than Dutch pronouns, presumably because pronouns in Turkish are pragmatically marked forms. We did not find any cross-linguistic influence in bilingual speech or gesture patterns, in line with studies (speech only) of highly proficient bilinguals. We therefore suggest that speech and gesture parallel each other not only in monolingual but also in bilingual production. Highly proficient heritage speakers who have been exposed to diverse linguistic and gestural patterns of each language from early on maintain monolingual patterns of pragmatic constraints on the use of pronouns multimodally.

Abstract

Recent work has made available a number of standardized meta- analyses bearing on various aspects of infant language processing. We utilize data from two such meta-analyses (discrimination of vowel contrasts and word segmentation, i.e., recognition of word forms extracted from running speech) to assess whether the published body of empirical evidence supports a bottom-up versus a top-down theory of early phonological development by leveling the power of results from thousands of infants. We predicted that if infants can rely purely on auditory experience to develop their phonological categories, then vowel discrimination and word segmentation should develop in parallel, with the latter being potentially lagged compared to the former. However, if infants crucially rely on word form information to build their phonological categories, then development at the word level must precede the acquisition of native sound categories. Our results do not support the latter prediction. We discuss potential implications and limitations, most saliently that word forms are only one top-down level proposed to affect phonological development, with other proposals suggesting that top-down pressures emerge from lexical (i.e., word-meaning pairs) development. This investigation also highlights general procedures by which standardized meta-analyses may be reused to answer theoretical questions spanning across phenomena.

Abstract

Theories of language acquisition and perceptual learning increasingly rely on statistical learning mechanisms. The current meta-analysis aims to clarify the robustness of this capacity in infancy within the word segmentation literature. Our analysis reveals a significant, small effect size for conceptual replications of Saffran, Aslin, & Newport (1996), and a nonsignificant effect across all studies that incorporate transitional probabilities to segment words. In both conceptual replications and the broader literature, however, statistical learning is moderated by whether stimuli are naturally produced or synthesized. These findings invite deeper questions about the complex factors that influence statistical learning, and the role of statistical learning in language acquisition.

Abstract

The current study investigates whether mentalizing, or taking the perspective of your interlocutor, plays an essential role throughout a conversation or whether it is mostly used in reaction to misunderstandings. This study is the first to use a brain-imaging method, MEG, to answer this question. In a first phase of the experiment, MEG participants interacted "live" with a confederate who set naming precedents for certain pictures. In a later phase, these precedents were sometimes broken by a speaker who named the same picture in a different way. This could be done by the same speaker, who set the precedent, or by a different speaker. Source analysis of MEG data showed that in the 800 ms before the naming, when the picture was already on the screen, episodic memory and language areas were activated, but no mentalizing areas, suggesting that the speaker's naming intentions were not anticipated by the listener on the basis of shared experiences. Mentalizing areas only became activated after the same speaker had broken a precedent, which we interpret as a reaction to the violation of conversational pragmatics.

Abstract

Rapid Automatized Naming (RAN) is a powerful tool for pre- dicting future reading skill. A person’s ability to quickly name symbols as they scan a table is related to higher-level reading proficiency in adults and is predictive of future literacy gains in children. However, noticeable differences are present in the strategies or patterns within groups having similar task comple- tion times. Thus, a further stratification of RAN dynamics may lead to better characterization and later intervention to support reading skill acquisition. In this work, we analyze the dynamics of the eyes, voice, and the coordination between the two during performance. It is shown that fast performers are more similar to each other than to slow performers in their patterns, but not vice versa. Further insights are provided about the patterns of more proficient subjects. For instance, fast performers tended to exhibit smoother behavior contours, suggesting a more sta- ble perception-production process.

Abstract

Brain oscillations have been shown to track the slow amplitude fluctuations in speech during comprehension. Moreover, there is evidence that these stimulus-induced cortical rhythms may persist even after the driving stimulus has ceased. However, how exactly this neural entrainment shapes speech perception remains debated. This behavioral study investigated whether and how the frequency and phase of an entrained rhythm would influence the temporal sampling of subsequent speech. In two behavioral experiments, participants were presented with slow and fast isochronous tone sequences, followed by Dutch target words ambiguous between as /ɑs/ “ash” (with a short vowel) and aas /a:s/ “bait” (with a long vowel). Target words were presented at various phases of the entrained rhythm. Both experiments revealed effects of the frequency of the tone sequence on target word perception: fast sequences biased listeners to more long /a:s/ responses. However, no evidence for phase effects could be discerned. These findings show that an entrained rhythm’s frequency, but not phase, influences the temporal sampling of subsequent speech. These outcomes are compatible with theories suggesting that sensory timing is evaluated relative to entrained frequency. Furthermore, they suggest that phase tracking of (syllabic) rhythms by theta oscillations plays a limited role in speech parsing.

Abstract

Speech is an acoustic signal with inherent amplitude modulations in the 1-9 Hz range. Recent models of speech perception propose that this rhythmic nature of speech is central to speech recognition. Moreover, rhythmic amplitude modulations have been shown to have beneficial effects on language processing and the subjective impression listeners have of the speaker. This study investigated the role of amplitude modulations in the political arena by comparing the speech produced by Hillary Clinton and Donald Trump in the three presidential debates of 2016. Inspection of the modulation spectra, revealing the spectral content of the two speakers’ amplitude envelopes after matching for overall intensity, showed considerably greater power in Clinton’s modulation spectra (compared to Trump’s) across the three debates, particularly in the 1-9 Hz range. The findings suggest that Clinton’s speech had a more pronounced temporal envelope with rhythmic amplitude modulations below 9 Hz, with a preference for modulations around 3 Hz. This may be taken as evidence for a more structured temporal organization of syllables in Clinton’s speech, potentially due to more frequent use of preplanned utterances. Outcomes are interpreted in light of the potential beneficial effects of a rhythmic temporal envelope on intelligibility and speaker perception.

Abstract

Lexically guided perceptual learni ng refers to the use of lexical knowledge to retune sp eech categories and thereby adapt to a novel talker’s pronunciation. This adaptation has been extensively documented, but primarily for segmental-based learning in English and Dutch. In languages with lexical tone, such as Mandarin Chinese, tonal categories can also be retuned in this way, but segmental category retuning had not been studied. We report two experiment s in which Mandarin Chinese listeners were exposed to an ambiguous mixture of [f] and [s] in lexical contexts favoring an interpretation as either [f] or [s]. Listeners were subsequently more likely to identify sounds along a continuum between [f] and [s], and to interpret minimal word pairs, in a manner consistent with this exposure. Thus lexically guided perceptual learning of segmental categories had indeed taken place, consistent with suggestions that such learning may be a universally available adaptation process

Abstract

In order to understand early word learning we need to better understand and quantify properties of the input that young children receive. We extended the human simulation paradigm (HSP) using egocentric videos taken from infant head-mounted cameras. The videos were further annotated with gaze information indicating in-the-moment visual attention from the infant. Our new HSP prompted participants for two types of responses, thus differentiating referential from semantic ambiguity in the learning input. Consistent with findings on visual attention in word learning, we find a strongly bimodal distribution over HSP accuracy. Even in this open-ended task, most videos only lead to a small handful of common responses. What's more, referential ambiguity was the key bottleneck to performance: participants can nearly always recover the exact word that was said if they identify the correct referent. Finally, analysis shows that adult learners relied on particular, multimodal behavioral cues to infer those target referents.

Abstract

Interoperable annotation formats are fundamental to the utility, expansion, and sustainability of collective data repositories.In language development research, shared annotation schemes have been critical to facilitating the transition from raw acoustic data to searchable, structured corpora. Current schemes typically require comprehensive and manual annotation of utterance boundaries and orthographic speech content, with an additional, optional range of tags of interest. These schemes have been enormously successful for datasets on the scale of dozens of recording hours but are untenable for long-format recording corpora, which routinely contain hundreds to thousands of audio hours. Long-format corpora would benefit greatly from (semi-)automated analyses, both on the earliest steps of annotation—voice activity detection, utterance segmentation, and speaker diarization—as well as later steps—e.g., classification-based codes such as child-vs-adult-directed speech, and speech recognition to produce phonetic/orthographic representations. We present an annotation workflow specifically designed for long-format corpora which can be tailored by individual researchers and which interfaces with the current dominant scheme for short-format recordings. The workflow allows semi-automated annotation and analyses at higher linguistic levels. We give one example of how the workflow has been successfully implemented in a large cross-database project.

Abstract

We investigate children’s online predictive processing as it occurs naturally, in conversation. We showed 1–7 year-olds short videos of improvised conversation between puppets, controlling for available linguistic information through phonetic manipulation. Even one- and two-year-old children made accurate and spontaneous predictions about when a turn-switch would occur: they gazed at the upcoming speaker before they heard a response begin. This predictive skill relies on both lexical and prosodic information together, and is not tied to either type of information alone. We suggest that children integrate prosodic, lexical, and visual information to effectively predict upcoming linguistic material in conversation.

Abstract

Child-directed speech is argued to facilitate language development, and is found cross-linguistically and cross-culturally to varying degrees. However, previous research has generally focused on short samples of child-caregiver interaction, often in the lab or with experimenters present. We test the generalizability of this phenomenon with an initial descriptive analysis of the speech heard by young children in a large, unique collection of naturalistic, daylong home recordings. Trained annotators coded automatically-detected adult speech 'utterances' from 61 homes across 4 North American cities, gathered from children (age 2-24 months) wearing audio recorders during a typical day. Coders marked the speaker gender (male/female) and intended addressee (child/adult), yielding 10,886 addressee and gender tags from 2,523 minutes of audio (cf. HB-CHAAC Interspeech ComParE challenge; Schuller et al., in press). Automated speaker-diarization (LENA) incorrectly gender-tagged 30% of male adult utterances, compared to manually-coded consensus. Furthermore, we find effects of SES and gender on child-directed and overall speech, increasing child-directed speech with child age, and interactions of speaker gender, child gender, and child age: female caretakers increased their child-directed speech more with age than male caretakers did, but only for male infants. Implications for language acquisition and existing classification algorithms are discussed.

Abstract

In this study, we analyze one French signing and one French speaking family’s interaction during dinner. The families composed of two parents and two children aged 3 to 11 were filmed with three cameras to capture all family members’ behaviors. The three videos per dinner were synchronized and coded on ELAN. We annotated all participants’ acting, and languaging.
Our quantitative analyses show how family members collaboratively manage multiple streams of activity through the embodied performances of dining and interacting. We uncover different profiles according to participants’ modality of expression and status (focusing on the mother and the younger child). The hearing participants’ co-activity management illustrates their monitoring of dining and conversing and how they progressively master the affordances of the visual and vocal channels to maintain the simultaneity of the two activities. The deaf mother skillfully manages to alternate smoothly between dining and interacting. The deaf younger child manifests how she is in the process of developing her skills to manage multi-activity. Our qualitative analyses focus on the ecology of visual-gestural and audio-vocal languaging in the context of co-activity according to language and participant. We open new perspectives on the management of gaze and body parts in multimodal languaging.

Abstract

Assimilation of place of articulation across a nasal and a following stop consonant is obligatory in Japanese, but not in Dutch. In four experiments the processing of assimilated forms by speakers of Japanese and Dutch was compared, using a task in which listeners blended pseudo-word pairs such as ranga-serupa. An assimilated blend of this pair would be rampa, an unassimilated blend rangpa. Japanese listeners produced significantly more assimilated than unassimilated forms, both with pseudo-Japanese and pseudo-Dutch materials, while Dutch listeners produced significantly more unassimilated than assimilated forms in each materials set. This suggests that Japanese listeners, whose native-language phonology involves obligatory assimilation constraints, represent the assimilated nasals in nasal-stop sequences as unmarked for place of articulation, while Dutch listeners, who are accustomed to hearing unassimilated forms, represent the same nasal segments as marked for place of articulation.

Abstract

The perceptual processing of speech is a constant interplay of multiple competing albeit convergent processes: acoustic input vs. higher-level representations, universal mechanisms vs. language-specific, veridical traces of speech experience vs. construction and activation of abstract representations. The present summary concerns the third of these issues. The ability to generalise across experience and to deal with resulting abstractions is the hallmark of human cognition, visible even in early infancy. In speech processing, abstract representations play a necessary role in both production and perception. New sorts of evidence are now informing our understanding of the breadth of this role.

Abstract

In English and Dutch, listeners entrain to prosodic contours to predict where focus will fall in an utterance. However, is this strategy universally available, even in languages with different phonological systems? In a phoneme detection experiment, we examined whether prosodic entrainment is also found in Mandarin Chinese, a tone language, where in principle the use of pitch for lexical identity may take precedence over the use of pitch cues to salience. Consistent with the results from Germanic languages, response times were facilitated when preceding intonation predicted accent on the target-bearing word. Acoustic analyses revealed greater F0 range in the preceding intonation of the predicted-accent sentences. These findings have implications for how universal and language-specific mechanisms interact in the processing of salience.

Abstract

Listeners process spoken language in ways which are adapted to the phonological structure of their native language. As a consequence, non-native speakers do not listen to a language in the same way as native speakers; moreover, listeners may use their native language listening procedures inappropriately with foreign input. With sufficient experience, however, it may be possible to inhibit this latter (counter-productive) behavior.

Abstract

Languages contain tens of thousands of words, but these are constructed from a tiny handful of phonetic elements. Consequently, words resemble one another, or can be embedded within one another, a coup stick snot with standing. me process of spoken-word recognition by human listeners involves activation of multiple word candidates consistent with the input, and direct competition between activated candidate words. Further, human listeners are sensitive, at an early, prelexical, stage of speeeh processing, to constraints on what could potentially be a word of the language.

Abstract

English and French listeners performed two tasks - click location and speeded click detection - with both English and French sentences, closely matched for syntactic and phonological structure. Clicks were located more accurately in open- than in closed-class words in both English and French; they were detected more rapidly in open- than in closed-class words in English, but not in French. The two listener groups produced the same pattern of responses, suggesting that higher-level linguistic processing was not involved in these tasks.

Abstract

The phoneme detection task is widely used in spoken word recognition research. Alphabetically literate participants, however, are more used to explicit representations of letters than of phonemes. The present study explored whether phoneme detection is sensitive to how target phonemes are, or may be, orthographically realised. Listeners detected the target sounds [b,m,t,f,s,k] in word-initial position in sequences of isolated English words. Response times were faster to the targets [b,m,t], which have consistent word-initial spelling, than to the targets [f,s,k], which are inconsistently spelled, but only when listeners’ attention was drawn to spelling by the presence in the experiment of many irregularly spelled fillers. Within the inconsistent targets [f,s,k], there was no significant difference between responses to targets in words with majority and minority spellings. We conclude that performance in the phoneme detection task is not necessarily sensitive to orthographic effects, but that salient orthographic manipulation can induce such sensitivity.

Abstract

The performance of automatic speech recognition systems is usually assessed in terms of error rate. Human speech recognition produces few errors, but relative difficulty of processing can be assessed via response time techniques. We report the construction of a measure analogous to response time in a machine recognition system. This measure may be compared directly with human response times. We conducted a trial comparison of this type at the phoneme level, including both tense and lax vowels and a variety of consonant classes. The results suggested similarities between human and machine processing in the case of consonants, but differences in the case of vowels.

Abstract

Word blends combine fragments from two words, either in speech errors or when a new word is created. Previous work has demonstrated that in Japanese, such blends preserve moraic structure; in English they do not. A similar effect of moraic structure is observed in perceptual research on segmentation of continuous speech in Japanese; English listeners, by contrast, exploit stress units in segmentation, suggesting that a general rhythmic constraint may underlie both findings. The present study examined whether mis parallel would also hold for word blends. In spontaneous English polysyllabic blends, the source words were significantly more likely to be split before a strong than before a weak (unstressed) syllable, i.e. to be split at a stress unit boundary. In an experiment in which listeners were asked to identify the source words of blends, significantly more correct detections resulted when splits had been made before strong syllables. Word blending, like speech segmentation, appears to be constrained by language rhythm.

Abstract

In a 50,000-word corpus of spoken British English the occurrence of words embedded within other words is reported. Within-word embedding in this real speech sample is common, and analogous to the extent of embedding observed in the vocabulary. Imposition of a syllable boundary matching constraint reduces but by no means eliminates spurious embedding. Embedded words are most likely to overlap with the beginning of matrix words, and thus may pose serious problems for speech recognisers.

Abstract

Vocabularies contain hundreds of thousands of words built from only a handful of phonemes, so that inevitably longer words tend to contain shorter ones. In many languages (but not all) such embedded words occur more often word-initially than word-finally, and this asymmetry, if present, has farreaching consequences for spoken-word recognition. Prior research had ascribed the asymmetry to suffixing or to effects of stress (in particular, final syllables containing the vowel schwa). Analyses of the standard French vocabulary here reveal an effect of suffixing, as predicted by this account, and further analyses of an artificial variety of French reveal that extensive final schwa has an independent and additive effect in promoting the embedding asymmetry.

Abstract

Spatial congruity effects have often been interpreted as evidence for metaphorical thinking, but an alternative markedness-based account challenges this view. In two experiments, we directly compared metaphor and markedness explanations for spatial congruity effects, using musical pitch as a testbed. English speakers who talk about pitch in terms of spatial height were tested in speeded space-pitch compatibility tasks. To determine whether space-pitch congruency effects could be elicited by any marked spatial continuum, participants were asked to classify high- and low-frequency pitches as 'high' and 'low' or as 'front' and 'back' (both pairs of terms constitute cases of marked continuums). We found congruency effects in high/low conditions but not in front/back conditions, indicating that markedness is not sufficient to account for congruity effects (Experiment 1). A second experiment showed that congruency effects were specific to spatial words that cued a vertical schema (tall/short), and that congruity effects were not an artifact of polysemy (e.g., 'high' referring both to space and pitch). Together, these results suggest that congruency effects reveal metaphorical uses of spatial schemas, not markedness effects.

Abstract

Responding to similarity, difference, and relative magnitude (SDM) is ubiquitous in the animal kingdom. However, humans seem unique in the ability to represent relative magnitude (‘more’/‘less’) and similarity (‘same’/‘different’) as abstract relations that take arguments (e.g., greater-than (x,y)). While many models use structured relational representations of magnitude and similarity, little progress has been made on how these representations arise. Models that developuse these representations assume access to computations of similarity and magnitude a priori, either encoded as features or as output of evaluation operators. We detail a mechanism for producing invariant responses to “same”, “different”, “more”, and “less” which can be exploited to compute similarity and magnitude as an evaluation operator. Using DORA (Doumas, Hummel, & Sandhofer, 2008), these invariant responses can serve be used to learn structured relational representations of relative magnitude and similarity from pixel images of simple shapes

Abstract

Humans readily generalize, applying prior knowledge to novel situations and stimuli. Advances in machine learning have begun to approximate and even surpass human performance, but these systems struggle to generalize what they have learned to untrained situations. We present a model based on wellestablished neurocomputational principles that demonstrates human-level generalisation. This model is trained to play one video game (Breakout) and performs one-shot generalisation to a new game (Pong) with different characteristics. The model
generalizes because it learns structured representations that are functionally symbolic (viz., a role-filler binding calculus) from unstructured training data. It does so without feedback, and without requiring that structured representations are specified a priori. Specifically, the model uses neural co-activation to discover which characteristics of the input are invariant and to learn relational predicates, and oscillatory regularities in network firing to bind predicates to arguments. To our knowledge,
this is the first demonstration of human-like generalisation in a machine system that does not assume structured representa-
tions to begin with.

Abstract

This paper summarizes the results of a descriptive syntactic category analysis of child English no which reveals that young children use and represent no as a determiner and negatives like no pen as NPs, contra standard analyses.

Abstract

In recent years, large scale initiatives like CLARIN set out to overcome the notorious heterogeneity of metadata formats in the domain of language resource. The CLARIN Component Metadata Infrastructure established means for flexible resouce descriptions for the domain of language resources. The Data Category Registry ISOcat and the accompanying Relation Registry foster semantic interoperability within the growing heterogeneous collection of metadata records. This paper describes the CMD Infrastructure focusing on the facilities for semantic mapping, and gives also an overview of the current status in the joint component metadata domain.

Abstract

We report the results of a large-scale (N=1571) experiment to investigate whether spoken words can emerge from the process of repeated imitation. Participants played a version of the children’s game “Telephone”. The first generation was asked to imitate recognizable environmental sounds (e.g., glass breaking, water splashing); subsequent generations imitated the imitators for a total of 8 generations. We then examined whether the vocal imitations became more stable and word-like, retained a resemblance to the original sound, and became more suitable as learned category labels. The results showed (1) the imitations became progressively more word-like, (2) even after 8 generations, they could be matched above chance to the environmental sound that motivated them, and (3) imitations from later generations were more effective as learned category labels. These results show how repeated imitation can create progressively more word-like forms while retaining a semblance of iconicity.

Abstract

The study presented here focuses on two pragmatic gestures:
the hand flip (Ferré, 2011), a gesture of the Palm Up Open
Hand/PUOH family (Müller, 2004) and the closed hand which
can be considered as the opposite kind of movement to the open-
ing of the hands present in the PUOH gesture. Whereas one of
the functions of the hand flip has been described as presenting
a new point in speech (Cienki, 2021), the closed hand gesture
has not yet been described in the literature to the best of our
knowledge. It can however be conceived of as having the oppo-
site function of announcing the end of a point in discourse. The
object of the present study is therefore to determine, with the
study of prosodic features, if the two gestures are found in the
same type of speech units and what their respective scope is.
Drawing from a corpus of three TED Talks in French the
prosodic characteristics of the speech that accompanies the two
gestures will be examined. The hypothesis developed in the
present paper is that their scope should be reflected in the
prosody of accompanying speech, especially pitch key, tone,
and relative pitch range. The prediction is that hand flips and
closing hand gestures are expected to be located at the periph-
ery of Intonation Phrases (IPs), Inter-Pausal Units (IPUs) or
more conversational Turn Constructional Units (TCUs), and are
likely to be co-occurrent with pauses in speech. But because of
the natural slope of intonation in speech, the speech that accom-
pany early gestures in Intonation Phrases should reveal different
features from the speech at the end of intonational units. Tones
should be different as well, considering the prosodic structure
of spoken French.

Abstract

One of the most daunting tasks of a listener is to map a
continuous auditory stream onto known speech sound
categories and lexical items. A major issue with this mapping
problem is the variability in the acoustic realizations of sound
categories, both within and across speakers. Past research has
suggested listeners may use visual information (e.g., lipreading)
to calibrate these speech categories to the current
speaker. Previous studies have focused on audiovisual
recalibration of consonant categories. The present study
explores whether vowel categorization, which is known to show
less sharply defined category boundaries, also benefit from
visual cues.
Participants were exposed to videos of a speaker
pronouncing one out of two vowels, paired with audio that was
ambiguous between the two vowels. After exposure, it was
found that participants had recalibrated their vowel categories.
In addition, individual variability in audiovisual recalibration is
discussed. It is suggested that listeners’ category sharpness may
be related to the weight they assign to visual information in
audiovisual speech perception. Specifically, listeners with less
sharp categories assign more weight to visual information
during audiovisual speech recognition.

Abstract

A crucial aspect of everyday conversational interactions is our ability to establish and maintain common ground. Understanding the relevant mechanisms involved in such social coordination remains an important challenge for cognitive science. While common ground is often discussed in very general terms, different contexts of interaction are likely to afford different coordination mechanisms. In this paper, we investigate the presence and relation of three mechanisms of social coordination – backchannels, interactive alignment and conversational repair – across free and task-oriented conversations. We find significant differences: task-oriented conversations involve higher presence of repair – restricted offers in particular – and backchannel, as well as a reduced level of lexical and syntactic alignment. We find that restricted repair is associated with lexical alignment and open repair with backchannels. Our findings highlight the need to explicitly assess several mechanisms at once and to investigate diverse activities to understand their role and relations.

Abstract

We conduct the first systematic comparison of automated semantic
annotation based on either the full-text or only on the title metadata
of documents. Apart from the prominent text classification baselines
kNN and SVM, we also compare recent techniques of Learning
to Rank and neural networks and revisit the traditional methods
logistic regression, Rocchio, and Naive Bayes. Across three of our
four datasets, the performance of the classifications using only titles
reaches over 90% of the quality compared to the performance when
using the full-text.

Abstract

We assess the suitability of word embeddings for practical information retrieval scenarios. Thus, we assume that users issue ad-hoc short queries where we return the first twenty retrieved documents after applying a boolean matching operation between the query and the documents. We compare the performance of several techniques that leverage word embeddings in the retrieval models to compute the similarity between the query and the documents, namely word centroid similarity, paragraph vectors, Word Mover’s distance, as well as our novel inverse document frequency (IDF) re-weighted word centroid similarity. We evaluate the performance using the ranking metrics mean average precision, mean reciprocal rank, and normalized discounted cumulative gain. Additionally, we inspect the retrieval models’ sensitivity to document length by using either only the title or the full-text of the documents for the retrieval task. We conclude that word centroid similarity is the best competitor to state-of-the-art retrieval models. It can be further improved by re-weighting the word frequencies with IDF before aggregating the respective word vectors of the embedding. The proposed cosine similarity of IDF re-weighted word vectors is competitive to the TF-IDF baseline and even outperforms it in case of the news domain with a relative percentage of 15%.

Abstract

The last few years have revealed that several species may share the building blocks of Musicality with humans. The recognition of these building blocks (e.g., rhythm, frequency variation) was a necessary impetus for a new round of studies investigating rhythmic variation in animal vocal displays. Singing primates are a small group of primate species that produce modulated songs ranging from tens to thousands of vocal units. Previous studies showed that the indri, the only singing lemur, is currently the only known species that perform duet and choruses showing multiple rhythmic categories, as seen in human music. Rhythmic categories occur when temporal intervals between note onsets are not uniformly distributed, and rhythms with a small integer ratio between these intervals are typical of human music. Besides indris, white-handed gibbons and three crested gibbon species showed a prominent rhythmic category corresponding to a single small integer ratio, isochrony. This study reviews previous evidence on the co-occurrence of rhythmic categories in primates and focuses on the prospects for a comparative, multimodal study of rhythmicity in this clade.

Abstract

We propose a Random-Forest based sign language identification system. The system uses low-level visual features and is based on the hypothesis that sign languages have varying distributions of phonemes (hand-shapes, locations and movements). We evaluated the system on two sign languages -- British SL and Greek SL, both taken from a publicly available corpus, called Dicta Sign Corpus. Achieved average F1 scores are about 95% - indicating that sign languages can be identified with high accuracy using only low-level visual features.

Abstract

We present a vision-based method for signer diarization -- the task of automatically determining "who signed when?" in a video. This task has similar motivations and applications as speaker diarization but has received little attention in the literature. In this paper, we motivate the problem and propose a method for solving it. The method is based on the hypothesis that signers make more movements than their interlocutors. Experiments on four videos (a total of 1.4 hours and each consisting of two signers) show the applicability of the method. The best diarization error rate (DER) obtained is 0.16.

Abstract

This paper presents a Native Language Identification (NLI) system based on TF-IDF weighting schemes and using linear classifiers - support vector machines, logistic regressions and perceptrons. The system was one of the participants of the 2013 NLI Shared Task in the closed-training track, achieving 0.814 overall accuracy for a set of 11 native languages. This accuracy was only 2.2 percentage points lower than the winner's performance. Furthermore, with subsequent evaluations using 10-fold cross-validation (as given by the organizers) on the combined training and development data, the best average accuracy obtained is 0.8455 and the features that contributed to this accuracy are the TF-IDF of the combined unigrams and bigrams of words.

Abstract

We present and solve the speaker diarization problem in a novel way. We hypothesize that the gesturer is the speaker and that identifying the gesturer can be taken as identifying the active speaker. We provide evidence in support of the hypothesis from gesture literature and audio-visual synchrony studies. We also present a vision-only diarization algorithm that relies on gestures (i.e. upper body movements). Experiments carried out on 8.9 hours of a publicly available dataset (the AMI meeting data) show that diarization error rates as low as 15% can be achieved.

Abstract

How are space and time related in the brain? This study contrasts two proposals that make different predictions about the interaction between spatial and temporal magnitudes. Whereas ATOM implies that space and time are symmetrically related, Metaphor Theory claims they are asymmetrically related. Here we investigated whether space and time activate the same neural structures in the inferior parietal cortex (IPC) and whether the activation is symmetric or asymmetric across domains. We measured participants’ neural activity while they made temporal and spatial judgments on the same visual stimuli. The behavioral results replicated earlier observations of a space-time asymmetry: Temporal judgments were more strongly influenced by irrelevant spatial information than vice versa. The BOLD fMRI data indicated that space and time activated overlapping clusters in the IPC and that, consistent with Metaphor Theory, this activation was asymmetric: The shared region of IPC was activated more strongly during temporal judgments than during spatial judgments. We consider three possible interpretations of this neural asymmetry, based on 3 possible functions of IPC.

Abstract

Despite the early emergence of pointing, children are generally not documented to produce iconic gestures until later in development. Although research has described this developmental trajectory and the types of iconic gestures that emerge first, there has been limited focus on iconic gestures within interactional contexts. This study identified the first 10 iconic gestures produced by five monolingual English-speaking children in a naturalistic longitudinal video corpus and analysed the interactional contexts. We found children produced their first iconic gesture between 12 and 20 months and that gestural types varied. Although 34% of gestures could have been imitated or derived from adult or child actions in the preceding context, the majority were produced independently of any observed model. In these cases, adults often led the interaction in a direction where iconic gesture was an appropriate response. Overall, we find infants can represent a referent symbolically and possess a greater capacity for innovation than previously assumed. In order to develop our understanding of how children learn to produce iconic gestures, it is important to consider the immediate interactional context. Conducting naturalistic corpus analyses could be a more ecologically valid approach to understanding how children learn to produce iconic gestures in real life contexts.

Abstract

The shape of pitch contours has been shown to have an effect on the perceived duration of vowels. For instance, vowels with high level pitch and vowels with falling contours sound longer than vowels with low level pitch. Depending on whether the
comparison is between level pitches or between level and dynamic contours, these findings have been interpreted in two ways. For inter-level comparisons, where the duration results are the reverse of production results, a hypercorrection strategy in production has been proposed [1]. By contrast, for comparisons between level pitches and dynamic contours, the
longer production data for dynamic contours have been held responsible. We report an experiment with Dutch and Chinese listeners which aimed to show that production data and perception data are each other’s opposites for high, low, falling and rising contours. We explain the results, which are consistent with earlier findings, in terms of the compensatory listening strategy of [2], arguing that the perception effects are due to a perceptual compensation of articulatory strategies and
constraints, rather than that differences in production compensate for psycho-acoustic perception effects.

Abstract

There is evidence that the process of executing a planned utterance involves the use of both preceding-context and top-down cues. Utterance-initial words are cued only by the top-down plan. In contrast, non-initial words are cued both by top-down cues and preceding-context cues. Co-existence of both cue types raises the question of how they interact during learning. We argue that this interaction is competitive: items that tend to be preceded by predictive preceding-context cues are harder to activate from the plan without this predictive context. A novel computational model of this competition is developed. The model is tested on a corpus of repetition disfluencies and shown to account for the influences on patterns of restarts during production. In particular, this model predicts a novel Initiation Effect: following an interruption, speakers re-initiate production from words that tend to occur in utterance-initial position, even when they are not initial in the interrupted utterance.

Abstract

The representations generated by many mod-
els of language (word embeddings, recurrent
neural networks and transformers) correlate
to brain activity recorded while people read.
However, these decoding results are usually
based on the brain’s reaction to syntactically
and semantically sound language stimuli. In
this study, we asked: how does an LSTM (long
short term memory) language model, trained
(by and large) on semantically and syntac-
tically intact language, represent a language
sample with degraded semantic or syntactic
information? Does the LSTM representation
still resemble the brain’s reaction? We found
that, even for some kinds of nonsensical lan-
guage, there is a statistically significant rela-
tionship between the brain’s activity and the
representations of an LSTM. This indicates
that, at least in some instances, LSTMs and the
human brain handle nonsensical data similarly.

Abstract

In human face-to-face communication, language comprehension is a multi-modal, situated activity. However, little is known about how we combine information from these different modalities, and how perceived communicative intentions, often signaled through visual signals, such as eye
gaze, may influence this processing. We address this question by simulating a triadic communication context in which a
speaker alternated her gaze between two different recipients. Participants thus viewed speech-only or speech+gesture
object-related utterances when being addressed (direct gaze) or unaddressed (averted gaze). Two object images followed
each message and participants’ task was to choose the object that matched the message. Unaddressed recipients responded significantly slower than addressees for speech-only
utterances. However, perceiving the same speech accompanied by gestures sped them up to a level identical to
that of addressees. That is, when speech processing suffers due to not being addressed, gesture processing remains intact and enhances the comprehension of a speaker’s message

Abstract

Models of cognitive processes often include simplifications, idealisations, and fictionalisations, so how should we learn about cognitive processes from such models? Particularly in cognitive science, when many features of the target system are unknown, it is not always clear which simplifications, idealisations, and so on, are appropriate for a research question, and which are highly misleading. Here we use a case-study from studies of language evolution, and ideas from philosophy of science, to illustrate a robustness approach to learning from models. Robust properties are those that arise across a range of models, simulations and experiments, and can be used to identify key causal structures in the models, and the phenomenon, under investigation. For example, in studies of language evolution, the emergence of compositional structure is a robust property across models, simulations and experiments of cultural transmission, but only under pressures for learnability and expressivity. This arguably illustrates the principles underlying real cases of language evolution. We provide an outline of the robustness approach, including its limitations, and suggest that this methodology can be productively used throughout cognitive science. Perhaps of most importance, it suggests that different modelling frameworks should be used as tools to identify the abstract properties of a system, rather than being definitive expressions of theories.

Abstract

Attempts to connect individual differences in statistical learning with broader aspects of cognition have received considerable attention, but have yielded mixed results. A possible explanation is that statistical learning is typically tested using the two-alternative forced choice (2AFC) task. As a meta-cognitive task relying on explicit familiarity judgments, 2AFC may not accurately capture implicitly formed statistical computations. In this paper, we adapt the classic serial-recall memory paradigm to implicitly test statistical learning in a statistically-induced chunking recall (SICR) task. We hypothesized that artificial language exposure would lead subjects to chunk recurring statistical patterns, facilitating recall of words from the input. Experiment 1 demonstrates that SICR offers more fine-grained insights into individual differences in statistical learning than 2AFC. Experiment 2 shows that SICR has higher test-retest reliability than that reported for 2AFC. Thus, SICR offers a more sensitive measure of individual differences, suggesting that basic chunking abilities may explain statistical learning.

Abstract

Most bioacoustics software is used to analyse the already collected acoustics data in batch, i.e., after the data-collecting phase of a scientific study. However, experiments based on animal training require immediate and precise reactions from the experimenter, and thus do not easily dovetail with a typical bioacoustics workflow. Bridging this methodological gap, we have developed a custom application to live-monitor the vocal development of harbour seals in a behavioural experiment. In each trial, the application records and automatically detects an animal's call, and immediately measures duration and acoustic measures such as intensity, fundamental frequency, or formant frequencies. It then displays a spectrogram of the recording and the acoustic measurements, allowing the experimenter to instantly evaluate whether or not to reinforce the animal's vocalisation. From a technical perspective, the rapid and easy development of this custom software was made possible by combining multiple open-source software projects. Here, we integrated the acoustic analyses from Parselmouth, a Python library for Praat, together with PyAudio and Matplotlib's recording and plotting functionality, into a custom graphical user interface created with PyQt. This flexible recombination of different open-source Python libraries allows the whole program to be written in a mere couple of hundred lines of code

Abstract

Second language (L2) research often involves analyses of acoustic measures of fluency. The studies investigating fluency, however, have been difficult to compare because the measures of fluency that were used differed widely. One of the differences between studies concerns the lower cut-off point for silent pauses, which has been set anywhere between 100 ms and 1000 ms. The goal of this paper is to find an optimal cut-off point. We calculate acoustic measures of fluency using different pause thresholds and then relate these measures to a measure of L2 proficiency and to ratings on fluency.

Abstract

Sign language is a rich form of communication, uniquely conveying meaning through a combination of gestures, facial expressions, and body movements. Existing research in sign language generation has predominantly focused on text-to-sign pose generation, while speech-to-sign pose generation remains relatively underexplored. Speech-to-sign language generation models can facilitate effective communication between the deaf and hearing communities. In this paper, we propose an architecture that utilises prosodic information from speech audio and semantic context from text to generate sign pose sequences. In our approach, we adopt a multi-tasking strategy that involves an additional task of predicting Facial Action Units (FAUs). FAUs capture the intricate facial muscle movements that play a crucial role in conveying specific facial expressions during sign language generation. We train our models on an existing Indian Sign language dataset that contains sign language videos with audio and text translations. To evaluate our models, we report Dynamic Time Warping (DTW) and Probability of Correct Keypoints (PCK) scores. We find that combining prosody and text as input, along with incorporating facial action unit prediction as an additional task, outperforms previous models in both DTW and PCK scores. We also discuss the challenges and limitations of speech-to-sign pose generation models to encourage future research in this domain. We release our models, results and code to foster reproducibility and encourage future research1.

Abstract

In production, frequently used words are preferentially extended to new, though related meanings. In comprehension, frequent exposure to a word instead makes the learner confident that all of the word’s legitimate uses have been experienced, resulting in an entrenched form-meaning mapping between the word and its experienced meaning(s). This results in a perception-production dissociation, where the forms speakers are most likely to map onto a novel meaning are precisely the forms that they believe can never be used that way. At first glance, this result challenges the idea of bidirectional form-meaning mappings, assumed by all current approaches to linguistic theory. In this paper, we show that bidirectional form-meaning mappings are not in fact challenged by this production-perception dissociation. We show that the production-perception dissociation is expected even if learners of the lexicon acquire simple symmetrical form-meaning associations through simple Hebbian learning.

Abstract

Deaf children born to hearing parents are exposed to language input quite late, which has long-lasting effects on language production. Previous studies with deaf individuals mostly focused on linguistic expressions of motion events, which have several event components. We do not know if similar effects emerge in simple events such as descriptions of spatial configurations of objects. Moreover, previous data mainly come from late adult signers. There is not much known about language development of late signing children soon after learning sign language. We compared simple event descriptions of late signers of Turkish Sign Language (adults, children) to age-matched native signers. Our results indicate that while late signers in both age groups are native-like in frequency of expressing a relational encoding, they lag behind native signers in using morphologically complex linguistic forms compared to other simple forms. Late signing children perform similar to adults and thus showed no development over time.

Abstract

English and German have similar prosody, but their speakers realize some pitch falls (not rises) in subtly different ways. We here test for asymmetry in perception. An ABX discrimination task requiring F0 slope or duration judgements on isolated vowels revealed no cross-language difference in duration or F0 fall discrimination, but discrimination of rises (realized similarly in each language) was less accurate for English than for German listeners. This unexpected finding may reflect greater sensitivity to rising patterns by German listeners, or reduced sensitivity by English listeners as a result of extensive exposure to phrase-final rises (“uptalk”) in their language

Abstract

Why do languages have the categories they do? It has been argued that spatial terms in the world’s languages reflect categories that support highly informative communication, and that this accounts for the spatial categories found across languages. However, this proposal has been tested against only nine languages, and in a limited fashion. Here, we consider two new languages: Maijɨki, an under-documented language of Peruvian Amazonia, and English. We analyze spatial data from these two new languages and the original nine, using thorough and theoretically targeted computational tests. The results support the hypothesis that spatial terms across dissimilar languages enable near-optimally informative communication, over an influential competing hypothesis

Abstract

A central question in the psycholinguistic study of multilingualism is how syntax is shared across languages. We implement a model to investigate whether error-based implicit learning can provide an account of cross-language structural priming. The model is based on the Dual-path model of
sentence-production (Chang, 2002). We implement our model using the Bilingual version of Dual-path (Tsoukala, Frank, & Broersma, 2017). We answer two main questions: (1) Can structural priming of active and passive constructions occur between English and Spanish in a bilingual version of the Dual-
path model? (2) Does cross-language priming differ quantitatively from within-language priming in this model? Our results show that cross-language priming does occur in the model. This finding adds to the viability of implicit learning as an account of structural priming in general and cross-language
structural priming specifically. Furthermore, we find that the within-language priming effect is somewhat stronger than the cross-language effect. In the context of mixed results from
behavioral studies, we interpret the latter finding as an indication that the difference between cross-language and within-
language priming is small and difficult to detect statistically.

Abstract

The previous literature has suggested that the hand movement in co-speech gestures and signs consists of a series of phases with qualitatively different dynamic characteristics. In this paper, we propose a syntagmatic rule system for movement phases that applies to both co-speech gestures and signs. Descriptive criteria for the rule system were developed for the analysis video-recorded continuous production of signs and gesture. It involves segmenting a stream of body movement into phases and identifying different phase types. Two human coders used the criteria to analyze signs and cospeech gestures that are produced in natural discourse. It was found that the criteria yielded good inter-coder reliability. These criteria can be used for the technology of automatic recognition of signs and co-speech gestures in order to segment continuous production and identify the potentially meaningbearing phase.

Abstract

Interpreting the meaning of co-speech gesture often involves
identifying a gesture’s ‘lexical affiliate’, the word or phrase to
which it most closely relates (Schegloff 1984). Though there is
work within gesture studies that resists this simplex mapping of
meaning from speech to gesture (e.g. de Ruiter 2000; Kendon
2014; Parrill 2008), including an evolving body of literature on
recurrent gesture and gesture families (e.g. Fricke et al. 2014; Müller 2017), it is still the lexical affiliate model that is most ap-
parent in formal linguistic models of multimodal meaning(e.g.
Alahverdzhieva et al. 2017; Lascarides and Stone 2009; Puste-
jovsky and Krishnaswamy 2021; Schlenker 2020). In this work,
I argue that the lexical affiliate should be carefully reconsidered
in the further development of such models.
In place of the lexical affiliate, I suggest a further shift
toward a frame-based, action schematic approach to gestural
meaning in line with that proposed in, for example, Parrill and
Sweetser (2004) and Müller (2017). To demonstrate the utility
of this approach I present three types of compositional gesture
sequences which I call spatial contrast, spatial embedding, and
cooperative abstract deixis. All three rely on gestural context,
rather than gesture-speech alignment, to convey interactive (i.e.
pragmatic) meaning. The centrality of gestural context to ges-
ture meaning in these examples demonstrates the necessity of
developing a model of gestural meaning independent of its in-
tegration with speech.

Abstract

Using real-time eye-movement measures, we asked how a fantastical discourse context competes with stored representations of semantic and world knowledge to influence children's and adults' moment-by-moment interpretation of a story. Seven-year- olds were less effective at bypassing stored semantic and world knowledge during real-time interpretation than adults. Nevertheless, an effect of discourse context on comprehension was still apparent.

Abstract

Research in body language with use of gesture recognition and speech analysis has gained much attention in the recent times, influencing disciplines related to image and speech processing.

This study aims to design the Media Query Language (MQL) (Lenkiewicz, et al. 2012) combined with the Linguistic Media Query Interface (LMQI) for Elan (Wittenburg, et al. 2006). The system integrated with the new achievements in audio-video recognition will allow querying media files with predefined gesture phases (or motion primitives) and speech characteristics as well as combinations of both. For the purpose of this work the predefined motions and speech characteristics are called patterns for atomic elements and actions for a sequence of patterns. The main assumption is that a user-customized library of patterns and actions and automated media annotation with LMQI will reduce annotation time, hence decreasing costs of creation of annotated corpora. Increase of the number of annotated data should influence the speed and number of possible research in disciplines in which human multimodal interaction is a subject of interest and where annotated corpora are required.

Abstract

Languages differ in the degree of semantic flexibility of their syntactic roles. For example, Eng-
lish and Indonesian are considered more flexible with regard to the semantics of subjects,
whereas German and Japanese are less flexible. In Hawkins’ classification, more flexible lan-
guages are said to have a loose fit, and less flexible ones are those that have a tight fit. This
classification has been based on manual inspection of example sentences. The present paper
proposes a new, quantitative approach to deriving the measures of looseness and tightness from
corpora. We use corpora of online news from the Leipzig Corpora Collection in thirty typolog-
ically and genealogically diverse languages and parse them syntactically with the help of the
Universal Dependencies annotation software. Next, we compute Mutual Information scores for
each language using the matrices of lexical lemmas and four syntactic dependencies (intransi-
tive subjects, transitive subject, objects and obliques). The new approach allows us not only to
reproduce the results of previous investigations, but also to extend the typology to new lan-
guages. We also demonstrate that verb-final languages tend to have a tighter relationship be-
tween lexemes and syntactic roles, which helps language users to recognize thematic roles early
during comprehension.

Abstract

Large language models that exhibit instruction-following behaviour represent one of the biggest recent upheavals in conversational interfaces, a trend in large part fuelled by the release of OpenAI's ChatGPT, a proprietary large language model for text generation fine-tuned through reinforcement learning from human feedback (LLM+RLHF). We review the risks of relying on proprietary software and survey the first crop of open-source projects of comparable architecture and functionality. The main contribution of this paper is to show that openness is differentiated, and to offer scientific documentation of degrees of openness in this fast-moving field. We evaluate projects in terms of openness of code, training data, model weights, RLHF data, licensing, scientific documentation, and access methods. We find that while there is a fast-growing list of projects billing themselves as 'open source', many inherit undocumented data of dubious legality, few share the all-important instruction-tuning (a key site where human labour is involved), and careful scientific documentation is exceedingly rare. Degrees of openness are relevant to fairness and accountability at all points, from data collection and curation to model architecture, and from training and fine-tuning to release and deployment.

Abstract

Speech recognition systems are a key intermediary in voice-driven human-computer interaction. Although speech recognition works well for pristine monologic audio, real-life use cases in open-ended interactive settings still present many challenges. We argue that timing is mission-critical for dialogue systems, and evaluate 5 major commercial ASR systems for their conversational and multilingual support. We find that word error rates for natural conversational data in 6 languages remain abysmal, and that overlap remains a key challenge (study 1). This impacts especially the recognition of conversational words (study 2), and in turn has dire consequences for downstream intent recognition (study 3). Our findings help to evaluate the current state of conversational ASR, contribute towards multidimensional error analysis and evaluation, and identify phenomena that need most attention on the way to build robust interactive speech technologies.

Abstract

Previous research has shown that repeated interactions can cause iconicity in signals to reduce. However, data from several recent studies has shown the opposite trend: an increase in iconicity as the result of repeated interactions. Here, we discuss whether signals may become less or more iconic as a result of the modality used to produce them. We review several recent experimental results before presenting new data from multi-modal signals, where visual input creates audio feedback. Our results show that the growth in iconicity present in the audio information may come at a cost to iconicity in the visual information. Our results have implications for how we think about and measure iconicity in artificial signalling experiments. Further, we discuss how iconicity in real world speech may stem from auditory, kinetic or visual information, but iconicity in these different modalities may conflict.

Abstract

Children learn language from the speech in their home environment. Recent work shows that more infant-directed speech
(IDS) leads to stronger lexical development. But what makes IDS a particularly useful learning signal? Here, we expand on an attention-based account first proposed by Räsänen et al. (2018): that prosodic modifications make IDS less predictable, and thus more interesting. First, we reproduce the critical finding from Räsänen et al.: that lab-recorded IDS pitch is less predictable compared to adult-directed speech (ADS). Next, we show that this result generalizes to the home language environment, finding that IDS in daylong recordings is also less predictable than ADS but that this pattern is much less robust than for IDS recorded in the lab. These results link experimental work on attention and prosodic modifications of IDS to real-world language-learning environments, highlighting some challenges of scaling up analyses of IDS to larger datasets that better capture children’s actual input.

Abstract

Lexical stress is realised similarly in English, German, and
Dutch. On a suprasegmental level, stressed syllables tend to be
longer and more acoustically salient than unstressed syllables;
segmentally, vowels in unstressed syllables are often reduced.
The frequency of unreduced unstressed syllables (where only
the suprasegmental cues indicate lack of stress) however,
differs across the languages. The present studies test whether
listener behaviour is affected by these vocabulary differences,
by investigating German listeners’ use of suprasegmental cues
to lexical stress in German and English word recognition. In a
forced-choice identification task, German listeners correctly
assigned single-syllable fragments (e.g., Kon-) to one of two
words differing in stress (KONto, konZEPT). Thus, German
listeners can exploit suprasegmental information for
identifying words. German listeners also performed above
chance in a similar task in English (with, e.g., DIver, diVERT),
i.e., their sensitivity to these cues also transferred to a nonnative
language. An English listener group, in contrast, failed
in the English fragment task. These findings mirror vocabulary
patterns: German has more words with unreduced unstressed
syllables than English does.

Abstract

Since the cognitive revolution, a widely held assumption has been that—whereas content may vary across cultures—cognitive processes would be universal, especially those on the more basic levels. Even if scholars do not fully subscribe to this assumption, they often conceptualize, or tend to investigate, cognition as if it were universal (Henrich, Heine, & Norenzayan, 2010). The insight that universality must not be presupposed but scrutinized is now gaining ground, and cognitive diversity has become one of the hot (and controversial) topics in the field (Norenzayan & Heine, 2005). We argue that, for scrutinizing the cultural dimension of cognition, taking an anthropological perspective is invaluable, not only for the task itself, but for attenuating the home-field disadvantages that are inescapably linked to cross-cultural research (Medin, Bennis, & Chandler, 2010).

Abstract

Speech rate is known to modulate perception of temporally ambiguous speech sounds. For instance, a vowel may be perceived as short when the immediate speech context is slow, but as long when the context is fast. Yet, effects of long-term tracking of speech rate are largely unexplored. Two experiments tested whether long-term tracking of rate influences perception of the temporal Dutch vowel contrast /ɑ/-/a:/. In Experiment 1, one low-rate group listened to 'neutral' rate speech from talker A and to slow speech from talker B. Another high-rate group was exposed to the same neutral speech from A, but to fast speech from B. Between-group comparison of the 'neutral' trials revealed that the low-rate group reported a higher proportion of /a:/ in A's 'neutral' speech, indicating that A sounded faster when B was slow. Experiment 2 tested whether one's own speech rate also contributes to effects of long-term tracking of rate. Here, talker B's speech was replaced by playback of participants' own fast or slow speech. No evidence was found that one's own voice affected perception of talker A in larger speech contexts. These results carry implications for our understanding of the mechanisms involved in rate-dependent speech perception and of dialogue.

Abstract

The results of a word-spotting experiment are presented in which Dutch listeners tried to spot different types of bisyllabic Dutch words embedded in different types of nonsense contexts. Embedded verbs were not reliably harder to spot than embedded nouns; this suggests that nouns and verbs are recognised via the same basic processes. Iambic words were no harder to spot than trochaic words, suggesting that trochaic words are not in principle easier to recognise than iambic words. Words were harder to spot in consonantal contexts (i.e., contexts which themselves could not be words) than in longer contexts which contained at least one vowel (i.e., contexts which, though not words, were possible words of Dutch). A control experiment showed that this difference was not due to acoustic differences between the words in each context. The results support the claim that spoken-word recognition is sensitive to the viability of sound sequences as possible words.

Abstract

This paper presents lexical statistics on the pattern of occurrence of words embedded in other words. We report the results of an analysis of 25000 words, varying in length from two to six syllables, extracted from a phonetically-coded English dictionary (The Longman Dictionary of Contemporary English). Each syllable, and each string of syllables within each word was checked against the dictionary. Two analyses are presented: the first used a complete list of polysyllables, with look-up on the entire dictionary; the second used a sublist of content words, counting only embedded words which were themselves content words. The results have important implications for models of human speech recognition. The efficiency of these models depends, in different ways, on the number and location of words within words.

Abstract

Learning how words refer to aspects of the environment is a complex task, but one that is supported by numerous cues within the environment which constrain the possibilities for matching words to their intended referents. In this paper we tested the predictions of a computational model of multiple cue integration for word learning, that predicted variation in the presence of cues provides an optimal learning situation. In a cross-situational learning task with adult participants, we varied the reliability of presence of distributional, prosodic, and gestural cues. We found that the best learning occurred when cues were often present, but not always. The effect of variability increased the salience of individual cues for the learner, but resulted in robust learning that was not vulnerable to individual cues’ presence or absence. Thus, variability of multiple cues in the language-learning environment provided the optimal circumstances for word learning.

Abstract

Many studies investigating the temporal alignment of co-speech
gestures to acoustic units in the speech signal find a close
coupling of the gestural landmarks and pitch accents or the
stressed syllable of pitch-accented words. In English, a pitch
accent is anchored in the lexically stressed syllable. Hence, it is
unclear whether it is the lexical phonological dimension of
stress, or the phrase-level prominence that determines the
details of speech-gesture synchronization. This paper explores
the relation between gestural phase transitions and accentual F0
peaks in Stockholm Swedish, which exhibits a lexical pitch
accent distinction. When produced with phrase-level
prominence, there are three different configurations of
lexicality of F0 peaks and the status of the syllable it is aligned
with. Through analyzing the alignment of the different F0 peaks
with gestural onsets in spontaneous dyadic conversations, we
aim to contribute to our understanding of the role of lexical
prosodic phonology in the co-production of speech and gesture.
The results, though limited by a small dataset, still suggest
differences between the three types of peaks concerning which
types of gesture phase onsets they tend to align with, and how
well these landmarks align with each other, although these
differences did not reach significance.