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

Infants learn the meaning of words from accumulated experiences of real-time interactions with their caregivers. To study the effects of visual sensory input on word learning, we recorded infant's view of the world using head-mounted eye trackers during free-flowing play with a caregiver. While playing, infants were exposed to novel label-object mappings and later learning outcomes for these items were tested after the play session. In this study we use a classification based approach to link properties of infants' visual scenes during naturalistic labeling moments to their word learning outcomes. We find that a model which integrates both highly informative and ambiguous sensory evidence is a better fit to infants' individual learning outcomes than models where either type of evidence is taken alone, and that raw labeling frequency is unable to account for the word learning differences we observe. Here we demonstrate how a computational model, using only raw pixels taken from the egocentric scene image, can derive insights on human language learning.

Abstract

In conversation, listeners frequently encounter speakers with foreign accents. Previous research on foreign-accented speech has primarily examined the short-term effects of exposure and the relative ease that listeners have with adapting to an accent. The present study examines the stability of this adaptation, with seven full days between testing sessions. On both days, subjects performed a cross-modal priming task in which they heard several minutes of an unfamiliar accent of their native language: a form of Hebrewaccented Dutch in which long /i:/ was shortened to /I/. During this task on Day 1, recognition of accented forms was not facilitated, compared to that of canonical forms. A week later, when tested on new words, facilitatory priming occurred, comparable to that seen for canonically produced items tested in both sessions. These results suggest that accented forms can be learned from brief exposure and the stable effects of this can be seen a week later.

Abstract

In the present paper we investigate the effect of categorising raw behavioural data or computational model responses. In addition, the effect of averaging over stimuli from potentially different populations is assessed. To this end, we replicate studies on word learning and generalisation abilities using the ACORNS models. Our results show that discrete categories may obscure interesting phenomena in the continuous responses. For example, the finding that learning in the model saturates very early at a uniform high recognition accuracy only holds for categorical representations. Additionally, a large difference in the accuracy for individual words is obscured by averaging over all stimuli. Because different words behaved differently for different speakers, we could not identify a phonetic basis for the differences. Implications and new predictions for infant behaviour are discussed.

Abstract

In the present paper we investigate the effect of categorising raw behavioural data or computational model responses. In addition, the effect of averaging over stimuli from potentially different populations is assessed. To this end, we replicate studies on word learning and generalisation abilities using the ACORNS models. Our results show that discrete
categories may obscure interesting phenomena in the continuous
responses. For example, the finding that learning in the model saturates very early at a uniform high recognition accuracy only holds for categorical representations. Additionally, a large difference in the accuracy for individual words is obscured
by averaging over all stimuli. Because different words behaved
differently for different speakers, we could not identify a phonetic
basis for the differences. Implications and new predictions for
infant behaviour are discussed.

Abstract

Studies have shown that cross-linguistically, consonants at the left edge of higher-level prosodic boundaries tend to be more forcefully articulated than those at lower-level boundaries, a phenomenon known as domain-initial strengthening. This study tests whether similar effects occur in Turkish, using the Autosegmental-Metrical model proposed by Ipek & Jun [1, 2] as the basis for assessing boundary strength. Productions of /t/ and /d/ were elicited in four domain-initial prosodic positions corresponding to progressively higher-level boundaries: syllable, word, intermediate phrase, and Intonational Phrase. A fifth position, nuclear word, was included in order to better situate it within the prosodic hierarchy. Acoustic correlates of articulatory strength were measured, including closure duration for /d/ and /t/, as well as voice onset time and burst energy for /t/. Our results show that closure duration increases cumulatively from syllable to intermediate phrase, while voice onset time and burst energy are not influenced by boundary strength. These findings provide corroborating evidence for Ipek & Jun’s model, particularly for the distinction between word and intermediate phrase boundaries. Additionally, articulatory strength at the left edge of the nuclear word patterned closely with word-initial position, supporting the view that the nuclear word is not associated with a distinct phrasing domain

Abstract

Space and time appear to be related asymmetrically in the child’s mind: temporal representations depend on spatial representations more than vice versa, as predicted by space-time metaphors in language. In a study supporting this conclusion, spatial information interfered with children’s temporal judgments more than vice versa (Casasanto, Fotakopoulou, & Boroditsky, 2010, Cognitive Science). In this earlier study, however, spatial information was available to participants for more time than temporal information was (as is often the case when people observe natural events), suggesting a skeptical explanation for the observed effect. Here we conducted a stronger test of the hypothesized space-time asymmetry, controlling spatial and temporal aspects of the stimuli even more stringently than they are generally ’controlled’ in the natural world. Results replicated Casasanto and colleagues’, validating their finding of a robust representational asymmetry between space and time, and extending it to children (4-10 y.o.) who speak Dutch and Brazilian Portuguese.

Abstract

Paper presented at The 161th Meeting Acoustical Society of America, Seattle, Washington, 23-27 May 2011. Reduction in casual speech can create ambiguity, e.g., “he was” can sound like “he’s.” Before quotative “like” “so she’s/she was like…”, it was found that there is little accurate acoustic information about the distinction in the signal. This work examines what types of information acoustics of the target itself, speech rate, coarticulation, and syntax/semantics listeners use to recognize such reduced function words. We compare perception studies presenting the targets auditorily with varying amounts of context, presenting the context without the targets, and a visual study presenting context in written form. Given primarily discourse information visual or auditory context only, subjects are strongly biased toward past, reflecting the use of quotative “like” for reporting past speech. However, if the target itself is presented, the direction of bias reverses, indicating that listeners favor acoustic information within the target which is reduced, sounding like the shorter, present form over almost any other source of information. Furthermore, when the target is presented auditorily with surrounding context, the bias shifts slightly toward the direction shown in the orthographic or auditory-no-target experiments. Thus, listeners prioritize acoustic information within the target when present, even if that information is misleading, but they also take discourse information into account.

Abstract

Listeners’ interactions often take place in auditorily challenging conditions. We examined how noise affects phonological competition during spoken word recognition. In a visual-world experiment, which allows us to examine the timecourse of recognition, English participants listened to target words in quiet and in noise while they saw four pictures on the screen: a target (e.g. candle), an onset overlap competitor (e.g. candy), an offset overlap competitor (e.g. sandal), and a distractor. The results showed that, while all competitors were relatively quickly suppressed in quiet listening conditions, listeners experienced persistent competition in noise from the offset competitor but not from the onset competitor. This suggests that listeners’ phonological competitor activation persists for longer in noise than in quiet and that listeners are able to deactivate some unwanted competition when listening to speech in noise. The well-attested competition pattern in quiet was not replicated. Possible methodological explanations for this result are discussed.

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

If concepts, categories, and word meanings are stable, how can people use them so flexibly? Here we explore a possible answer: maybe this stability is an illusion. Perhaps all concepts, categories, and word meanings (CC&Ms) are constructed ad hoc, each time we use them. On this proposal, all words are infinitely polysemous, all communication is ’good enough’, and no idea is ever the same twice. The details of people’s ad hoc CC&Ms are determined by the way retrieval cues interact with the physical, social, and linguistic context. We argue that even the most stable-seeming CC&Ms are instantiated via the same processes as those that are more obviously ad hoc, and vary (a) from one microsecond to the next within a given instantiation, (b) from one instantiation to the next within an individual, and (c) from person to person and group to group as a function of people’s experiential history. 826

Abstract

Can simple motor actions help people expand their vocabulary? Here we show that word learning depends on where students place their flash cards after studying them. In Experiment 1, participants learned the definitions of ”alien words” with positive or negative emotional valence. After studying each card, they placed it in one of two boxes (top or bottom), according to its valence. Participants who were instructed to place positive cards in the top box, consistent with Good is Up metaphors, scored about 10.

Abstract

When we think of children acquiring language, we often think of their acquisition of linguistic structure as separate from their acquisition of knowledge about the world. But it is clear that in the process of learning about language, children consult what they know about the world; and that in learning about the world, children use linguistic cues to discover how items are related to one another. This interaction between the acquisition of linguistic structure and the acquisition of category structure is especially clear in word learning.

Abstract

Pitch accents are analysed differently in an onramp analysis (i.e. ToBI) and an off-ramp analysis (e.g. Transcription of Dutch intonation - ToDI), two competing approaches in the Autosegmental Metrical tradition. A case in point is pre-final high rise. A pre-final rise is analysed as H* in ToBI but is phonologically ambiguous between H* or H*L (a (rise-)fall) in ToDI. This is because in ToDI, the L tone of a pre-final H*L can be realised in the following unaccented words and both H* and H*L can show up as a high rise in the accented word. To find out whether there is a two-way phonological contrast in pre-final high rises in Dutch, we examined the distribution of phonologically ambiguous high rises (H*(L)) and their phonetic realisation in different information structural conditions (topic vs. focus), compared to phonologically unambiguous H* and H*L. Results showed that there is indeed a H*L vs. H* contrast in prefinal high rises in Dutch and that H*L is realised as H*(L) when sonorant material is limited in the accented word. These findings provide new evidence for an off-ramp analysis of Dutch intonation and have far-reaching implications for analysis of intonation across languages.

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

In this study, we compared the performance of a long short-term memory (LSTM) neural network to the behavior of human participants on a language task that requires hierarchically structured knowledge. We show that humans interpret ambiguous noun phrases, such as second blue ball, in line with their hierarchical constituent structure. LSTMs, instead, only do
so after unambiguous training, and they do not systematically generalize to novel items. Overall, the results of our simulations indicate that a model can behave hierarchically without relying on hierarchical constituent structure.

Abstract

Previous research has shown that listeners use the prosodic structure of utterances in a predictive fashion in sentence comprehension, to direct attention to accented words. Acoustically identical words spliced into sentence contexts arc responded to differently if the prosodic structure of the context is \ aricd: when the preceding prosody indicates that the word will he accented, responses are faster than when the preceding prosodv is inconsistent with accent occurring on that word. In the present series of experiments speech hybridisation techniques were first used to interchange the timing patterns within pairs of prosodic variants of utterances, independently of the pitch and intensity contours. The time-adjusted utterances could then serve as a basis lor the orthogonal manipulation of the three prosodic dimensions of pilch, intensity and rhythm. The overall pattern of results showed that when listeners use prosody to predict accent location, they do not simply rely on a single prosodic dimension, hut exploit the interaction between pitch, intensity and rhythm.

Abstract

Listeners performed speeded two-alternative choice between two unfamiliar and relatively similar voices or between two phonetically close segments, in VC syllables. For each decision type (segment, voice), the non-target dimension (voice, segment) either was constant, or varied across four alternatives. Responses were always slower when a non-target dimension varied than when it did not, but the effect of phonetic variation on voice identity decision was stronger than that of voice variation on phonetic identity decision. Cues to voice and segment identity in speech are processed inter-dependently, but hard categorization decisions about voices draw on, and are hence sensitive to, segmental information.

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

One of a listener's major tasks in understanding continuous speech is segmenting the speech signal into separate words. When listening conditions are difficult, speakers can help listeners by deliberately speaking more clearly. In three experiments, we examined how word boundaries are produced in deliberately clear speech. We found that speakers do indeed attempt to mark word boundaries; moreover, they differentiate between word boundaries in a way which suggests they are sensitive to listener needs. Application of heuristic segmentation strategies makes word boundaries before strong syllables easiest for listeners to perceive; but under difficult listening conditions speakers pay more attention to marking word boundaries before weak syllables, i.e. they mark those boundaries which are otherwise particularly hard to perceive.

Abstract

Two lexical decision experiments in Cantonese are described in which the recognition of spoken target words as a function of phonological similarity to a preceding prime is investigated. Phonological similaritv in first syllables produced inhibition, while similarity in second syllables led to facilitation. Differences between syllables in tonal and segmental structure had generally similar effects.

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

Anaphora, a ubiquitous feature of natural language, poses a particular challenge to young children as they first learn language due to its referential ambiguity. In spite of this, parents and caregivers use anaphora frequently in child-directed speech, potentially presenting a risk to effective communication if children do not yet have the linguistic capabilities of resolving anaphora successfully. Through an eye-tracking study in a naturalistic free-play context, we examine the strategies that parents employ to calibrate their use of anaphora to their child's linguistic development level. We show that, in this way, parents are able to intuitively scaffold the complexity of their speech such that greater referential ambiguity does not hurt overall communication success.

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

Language acquisition involves learning nonadjacent dependencies that can obtain between words in a sentence. Several artificial grammar learning studies have shown that the ability of adults and children to detect dependencies between A and B in frames AXB is influenced by the amount of variation in the X element. This paper presents a model of statistical learning which displays similar behavior on this task and generalizes in a human-like way. The model was also used to predict human behavior for increased distance and more variation in dependencies. We compare our model-based approach with the standard invariance account of the variability effect.

Abstract

Graph neural networks (GNNs) have emerged as the standard method for numerous tasks on graph-structured data such as node classification. However, real-world graphs are often evolving over time and even new classes may arise. We model these challenges as an instance of lifelong learning, in which a learner faces a sequence of tasks and may take over knowledge acquired in past tasks. Such knowledge may be stored explicitly as historic data or implicitly within model parameters. In this work, we systematically analyze the influence of implicit and explicit knowledge. Therefore, we present an incremental training method for lifelong learning on graphs and introduce a new measure based on k-neighborhood time differences to address variances in the historic data. We apply our training method to five representative GNN architectures and evaluate them on three new lifelong node classification datasets. Our results show that no more than 50% of the GNN's receptive field is necessary to retain at least 95% accuracy compared to training over the complete history of the graph data. Furthermore, our experiments confirm that implicit knowledge becomes more important when fewer explicit knowledge is available.

Abstract

COVID-19 research datasets are crucial for analyzing research dynamics. Most collections of COVID-19 research items do not to include cited works and do not have annotations
from a controlled vocabulary. Starting with ZB MED KE data on COVID-19, which comprises CORD-19, we assemble a new dataset that includes cited work and MeSH annotations for all records. Furthermore, we conduct experiments on the analysis of research dynamics, in which we investigate predicting links in a co-annotation graph created on the basis of the new dataset. Surprisingly, we find that simple heuristic methods are better at
predicting future links than more sophisticated approaches such as graph neural networks.

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

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

Face to face interaction is fundamental to human sociality but is very complex to study in a scientific fashion. This theme issue brings together cutting-edge approaches to the study of face-to-face interaction and showcases how we can make progress in this area. Researchers are now studying interaction in adult conversation, parent-child relationships, neurodiverse groups, interactions with virtual agents and various animal species. The theme issue reveals how new paradigms are leading to more ecologically grounded and comprehensive insights into what social interaction is. Scientific advances in this area can lead to improvements in education and therapy, better understanding of neurodiversity and more engaging artificial agents

Abstract

The present paper considers the problem of annotating bibliographical references with labels/classes, given training data of references already annotated with labels. The problem is an instance of document categorization where the documents are short and written in a wide variety of languages. The skewed distributions of title words and labels calls for special carefulness when choosing a Machine Learning approach. The present paper describes how to induce Disjunctive Normal Form formulae (DNFs), which have several advantages over Decision Trees. The approach is evaluated on a large real-world collection of bibliographical references.

Abstract

This corpus study demonstrates that the realization of wordfinal /t/ in Dutch past-participles in various speech styles is affected by a word’s predictability and paradigmatic relative frequency. In particular, /t/s are shorter and more often absent if the two preceding words are more predictable. In addition, /t/s, especially in irregular verbs, are more reduced, the lower the verb’s lemma frequency relative to the past-participle’s frequency. Both effects are more pronounced in more spontaneous speech. These findings are expected if speech planning plays an important role in speech reduction. Index Terms: pronunciation variation, acoustic reduction, corpus research, word predictability, morphological decomposability

Abstract

Children with developmental language disorder (DLD) regularly use the base form of verbs (e.g., dance) instead of inflected forms (e.g., danced). We propose an account of this behavior in which children with DLD have difficulty processing novel inflected verbs in their input. This leads the inflected form to face stronger competition from alternatives. Competition is resolved by the production of a more accessible alternative with high semantic overlap with the inflected form: in English, the bare form. We test our account computationally by training a nonparametric Bayesian model that infers the productivity of the inflectional suffix (-ed). We systematically vary the number of novel types of inflected verbs in the input to simulate the input as processed by children with and without DLD. Modeling results are consistent with our hypothesis, suggesting that children’s inconsistent use of inflectional morphemes could stem from inferences they make on the basis of impoverished data.

Abstract

This special publication aims to build a renewed enthusiasm for collecting acquisition data across many languages, including those facing endangerment and loss. It presents a guide for documenting and describing child language and child-directed language in diverse languages and cultures, as well as a collection of acquisition sketches based on this guide. The guide is intended for anyone interested in working across child language and language documentation, including, for example, field linguists and language documenters, community language workers, child language researchers or graduate students.

Abstract

Using the visual-word paradigm, the present study investigated the effects of word onset and offset masking on the time course of non-native spoken-word recognition in the presence of background noise. In two experiments, Dutch non-native listeners heard English target words, preceded by carrier sentences that were noise-free (Experiment 1) or contained intermittent noise (Experiment 2). Target words were either onset- or offset-masked or not masked at all. Results showed that onset masking delayed target word recognition more than offset masking did, suggesting that – similar to natives – non-native listeners strongly rely on word onset information during word recognition in noise.

Abstract

This study uses a classical referential communication task to
investigate the role of co-speech gestures in the process of
coordination. The study manipulates both the common ground between the interlocutors, as well as the visibility of the gestures they use. The findings show that co-speech gestures are an integral part of the referential utterances speakers
produced with regard to both initial references as well as repeated references, and that the availability of gestures appears to impact on interlocutors’ referential oordination. The results are discussed with regard to past research on
common ground as well as theories of gesture production.

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

We investigated whether seeing a pointing gesture influences the perceived lexical stress. A pitch contour continuum between the Dutch words “CAnon” (‘canon’) and “kaNON” (‘cannon’) was presented along with a pointing gesture during the first or the second syllable. Pointing gestures following natural recordings but not Gaussian functions influenced stress perception (Experiment 1 and 2), especially when auditory context preceded (Experiment 2). This was not replicated in Experiment 3. Natural pointing gestures failed to affect the categorization of a pitch peak timing continuum (Experiment 4). There is thus no convincing evidence that seeing a pointing gesture influences lexical stress perception.

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

There is a strong relation between children’s exposure to
spatial terms and their later memory accuracy. In the current
study, we tested whether the production of spatial terms by
children themselves predicts memory accuracy and whether
and how language modality of these encodings modulates
memory accuracy differently. Hearing child speakers of
Turkish and deaf child signers of Turkish Sign Language
described pictures of objects in various spatial relations to each
other and later tested for their memory accuracy of these
pictures in a surprise memory task. We found that having
described the spatial relation between the objects predicted
better memory accuracy. However, the modality of these
descriptions in sign, speech, or speech-plus-gesture did not
reveal differences in memory accuracy. We discuss the
implications of these findings for the relation between spatial
language, memory, and the modality of encoding.

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

Annotations of media recordings are the grounds for linguistic research. Since creating those annotations is a very laborious task, reaching 100 times longer than the length of the annotated media, innovative audio and video processing algorithms are needed, in order to improve the efficiency and quality of annotation process. The AVATecH project, started by the Max-Planck Institute for Psycholinguistics (MPI) and the Fraunhofer institutes HHI and IAIS, aims at significantly speeding up the process of creating annotations of audio-visual data for humanities research. In order for this to be achieved a range of state-of-the-art audio and video pattern recognition algorithms have been developed and integrated into widely used ELAN annotation tool. To address the problem of heterogeneous annotation tasks and recordings we provide modular components extended by adaptation and feedback mechanisms to achieve competitive annotation quality within significantly less annotation time.

Abstract

Evolution and changes of all modern languages is a well-known fact. However, recently it is reaching dynamics never seen before, which results in loss of the vast amount of information encoded in every language. In order to preserve such heritage, properly annotated recordings of world languages are necessary. Since creating those annotations is a very laborious task, reaching times 100 longer than the length of the annotated media, innovative video processing algorithms are needed, in order to improve the efficiency and quality of annotation process.

Abstract

Processing medical data has always been an interesting field that has shown the need for effective image segmentation methods. Modern medical image segmentation solutions are focused on 3D image volumes, which originate at advanced acquisition devices. Operating on such data in a 3D envi- ronment is essential in order to take the full advantage of the available information. In this paper we present an extended version of our 3D image segmentation and reconstruction model that belongs to the family of Deformable Models and is capable of processing large image volumes in competitive times and in fully 3D environment, offering a big level of automation of the process and a high precision of results. It is also capable of handling topology changes and offers a very good scalability on multi-processing unit architectures. We present a description of the model and show its capabilities in the field of medical image processing.

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

How are the senses structured by the languages we speak, the cultures we inhabit? To what extent is the encoding of perceptual experiences in languages a matter of how the mind/brain is ―wired-up‖ and to what extent is it a question of local cultural preoccupation? The ―Language of Perception‖ project tests the hypothesis that some perceptual domains may be more ―ineffable‖ – i.e. difficult or impossible to put into words – than others. While cognitive scientists have assumed that proximate senses (olfaction, taste, touch) are more ineffable than distal senses (vision, hearing), anthropologists have illustrated the exquisite variation and elaboration the senses achieve in different cultural milieus. The project is designed to test whether the proximate senses are universally ineffable – suggesting an architectural constraint on cognition – or whether they are just accidentally so in Indo-European languages, so expanding the role of cultural interests and preoccupations. To address this question, a standardized set of stimuli of color patches, geometric shapes, simple sounds, tactile textures, smells and tastes have been used to elicit descriptions from speakers of more than twenty languages—including three sign languages. The languages are typologically, genetically and geographically diverse, representing a wide-range of cultures. The communities sampled vary in subsistence modes (hunter-gatherer to industrial), ecological zones (rainforest jungle to desert), dwelling types (rural and urban), and various other parameters. We examine how codable the different sensory modalities are by comparing how consistent speakers are in how they describe the materials in each modality. Our current analyses suggest that taste may, in fact, be the most codable sensorial domain across languages. Moreover, we have identified exquisite elaboration in the olfactory domains in some cultural settings, contrary to some contemporary predictions within the cognitive sciences. These results suggest that differential codability may be at least partly the result of cultural preoccupation. This shows that the senses are not just physiological phenomena but are constructed through linguistic, cultural and social practices.

Abstract

To study concepts, cognitive scientists must first identify some. The prevailing assumption is that they are revealed by words such as triangle, table, and robin. But languages vary dramatically in how they carve up the world by name. Either ordinary concepts must be heavily language-dependent or names cannot be a direct route to concepts. We asked English, Dutch, Spanish, and Japanese speakers to name videos of human locomotion and judge their similarities. We investigated what name inventories and scaling solutions on name similarity and on physical similarity for the groups individually and together suggest about the underlying concepts. Aggregated naming and similarity solutions converged on results distinct from the answers suggested by the word inventories and scaling solutions of any single language. Words such as triangle, table, and robin can help identify the conceptual space of a domain, but they do not directly reveal units of knowledge usefully considered 'concepts'.

Abstract

Visual and auditory channels have different affordances and
this is mirrored in what information is available for linguistic
encoding. The visual channel has high spatial acuity, whereas
the auditory channel has better temporal acuity. These
differences may lead to different conceptualizations of events
and affect multimodal language production. Previous studies of
motion events typically present visual input to elicit speech and
gesture. The present study compared events presented as audio-
only, visual-only, or multimodal (visual+audio) input and
assessed speech and co-speech gesture for path and manner of
motion in Turkish. Speakers with audio-only input mentioned
path more and manner less in verbal descriptions, compared to
speakers who had visual input. There was no difference in the
type or frequency of gestures across conditions, and gestures
were dominated by path-only gestures. This suggests that input
modality influences speakers’ encoding of path and manner of
motion events in speech, but not in co-speech gestures.

Abstract

How listeners understand spoken words despite massive variation in the speech signal is a central issue for linguistic theory. A recent focus on lexical frequency and specificity has proved fruitful in accounting for this phenomenon. Speech perception, though, is a multi-faceted process and likely incorporates a number of mechanisms to map a variable signal to meaning. We examine a well-established language use factor — lexical frequency — and how this factor is integrated with phonetic variability during the perception of accented speech. We show that an integrated perspective highlights a low-level perceptual mechanism that accounts for the perception of accented speech absent native contrasts, while shedding light on the use of interactive language factors in the perception of spoken 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

Recurrent neural networks (RNNs) have long been an architecture of interest for computational models of human sentence processing. The recently introduced Transformer architecture outperforms RNNs on many natural language processing tasks but little is known about its ability to model human language processing. We compare Transformer- and RNN-based language models’ ability to account for measures of human reading effort. Our analysis shows Transformers to outperform RNNs in explaining self-paced reading times and neural activity during reading English sentences, challenging the widely held idea that human sentence processing involves recurrent and immediate processing and provides evidence for cue-based retrieval.

Abstract

This study addresses the question whether visually grounded speech recognition (VGS) models learn to capture sentence semantics without access to any prior linguistic knowledge. We produce synthetic and natural spoken versions of a well known semantic textual similarity database and show that our VGS model produces embeddings that correlate well with human semantic similarity judgements. Our results show that a model trained on a small image-caption database outperforms two models trained on much larger databases, indicating that database size is not all that matters. We also investigate the importance of having multiple captions per image and find that this is indeed helpful even if the total number of images is lower, suggesting that paraphrasing is a valuable learning signal. While the general trend in the field is to create ever larger datasets to train models on, our findings indicate other characteristics of the database can just as important.

Abstract

Speakers tend to take over the articulatory habits of their interlocutors [e.g., Pardo, JASA (2006)]. This phonetic alignment could be the consequence of either a social mechanism or a direct and automatic link between speech perception and production. The latter assumes that social variables should have little influence on phonetic alignment. To test that participants were engaged in a "cloze task" (i.e., Stimulus: "In fantasy movies, silver bullets are used to kill ..." Response: "werewolves") with either one or four interlocutors. Given findings with the Asch-conformity paradigm in social psychology, multiple consistent speakers should exert a stronger force on the participant to align. To control the speech style of the interlocutors, their questions and answers were pre-recorded in either a formal or a casual speech style. The stimuli's speech style was then manipulated between participants and was consistent throughout the experiment for a given participant. Surprisingly, participants aligned less with the speech style if there were multiple interlocutors. This may reflect a "diffusion of responsibility:" Participants may find it more important to align when they interact with only one person than with a larger group.

Abstract

Language emergence is characterized by a high degree of lex-
ical variation. It has been suggested that the speed at which
lexical conventionalization occurs depends partially on social
structure. In large communities, individuals receive input from
many sources, creating a pressure for lexical convergence.
In small, insular communities, individuals can remember id-
iolects and share common ground with interlocuters, allow-
ing these communities to retain a high degree of lexical vari-
ation. We look at lexical variation in Kata Kolok, a sign lan-
guage which emerged six generations ago in a Balinese vil-
lage, where women tend to have more tightly-knit social net-
works than men. We test if there are differing degrees of lexical
uniformity between women and men by reanalyzing a picture
description task in Kata Kolok. We find that women’s produc-
tions exhibit less lexical uniformity than men’s. One possible
explanation of this finding is that women’s more tightly-knit
social networks allow for remembering idiolects, alleviating
the pressure for lexical convergence, but social network data
from the Kata Kolok community is needed to support this ex-
planation.

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.

Abstract

This paper describes the Glottolog/Langdoc project, an at- tempt to provide near-total bibliographical coverage of descriptive re- sources to the world's languages. Every reference is treated as a resource, as is every \languoid"[1]. References are linked to the languoids which they describe, and languoids are linked to the references described by them. Family relations between languoids are modeled in SKOS, as are relations across dierent classications of the same languages. This setup allows the representation of languoids as collections of references, render- ing the question of the denition of entities like `Scots', `West-Germanic' or `Indo-European' more empirical.

Abstract

We report two experiments in which vowels and steady-state consonants served as targets in a speeded detection task. In the first experiment, two vowels were compared with one voiced and once unvoiced fricative. Response times (RTs) to the vowels were longer than to the fricatives. The error rate was higher for the consonants. Consonants in word-final position produced the shortest RTs, For the vowels, RT correlated negatively with target duration. In the second experiment, the same two vowel targets were compared with two nasals. This time there was no significant difference in RTs, but the error rate was still significantly higher for the consonants. Error rate and length correlated negatively for the vowels only. We conclude that RT differences between phonemes are independent of vocalic or consonantal status. Instead, we argue that the process of phoneme detection reflects more finely grained differences in acoustic/articulatory structure within the phonemic repertoire.

Abstract

Phonetic convergence—i.e., adapting one’s speech
towards that of an interlocutor—has been shown
to occur in human-human conversations as well as
human-machine interactions. Here, we investigate
the hypothesis that human-to-robot convergence is
influenced by the human’s perception of the robot
and by the conversation’s topic. We conducted a
within-subjects experiment in which 33 participants
interacted with two robots differing in their eye gaze
behavior—one looked constantly at the participant;
the other produced gaze aversions, similarly to a
human’s behavior. Additionally, the robot asked
questions with increasing intimacy levels.
We observed that the speakers tended to converge
on F0 to the robots. However, this convergence
to the robots was not modulated by how the
speakers perceived them or by the topic’s intimacy.
Interestingly, speakers produced lower F0 means
when talking about more intimate topics. We
discuss these findings in terms of current theories of
conversational convergence.

Abstract

Japanese, British English and American English listeners were presented with spoken words in their native language, and asked to mark on a written transcript of each word the first natural division point in the word. The results showed clear and strong patterns of consensus, indicating that listeners have available to them conscious representations of within-word structure. Orthography did not play a strongly deciding role in the results. The patterns of response were at variance with results from on-line studies of speech segmentation, suggesting that the present task taps not those representations used in on-line listening, but levels of representation which may involve much richer knowledge of word-internal structure.

Abstract

Two groups of participants were trained to perceive an ambiguous sound [s/f] as either /s/ or /f/ based on lexical bias: One group heard the ambiguous fricative in /s/-final words, the other in /f/-final words. This kind of exposure leads to a recalibration of the /s/-/f/ contrast [e.g., 4]. In order to investigate when and how this recalibration emerges, test trials were interspersed among training and filler trials. The learning effect needed at least 10 clear training items to arise. Its emergence seemed to occur in a rather step-wise fashion. Learning did not improve much after it first appeared. It is likely, however, that the early test trials attracted participants' attention and therefore may have interfered with the learning process.