Search

O fenótipo ampliado do autismo (FAA) tem sido definido como traços subclínicos que configuram a expressão fenotípica de uma suscetibilidade genética para o desenvolvimento do Transtorno do Espectro Autista (TEA). A literatura tem apontado para a relação entre habilidades comunicativo-pragmáticas e autismo. No entanto, há controvérsia quanto à possibilidade dessas habilidades constituírem uma especificidade do FAA. Ainda, pesquisas com parentes de primeiro grau de pessoas com TEA ainda não obtiveram sucesso em estabelecer marcadores definitivos. Assim, este estudo comparou o desempenho de pais e mães de crianças com TEA (n=26) e sem TEA (n=26) em tarefas de habilidades comunicativo-pragmáticas e indicadores de FAA. Foram aplicados o Broader Autism Phenotype Questionnaire (BAPQ) e subtestes da Bateria Montreal de Avaliação da Comunicação (Bateria mac), além da Escala de Inteligência Wechsler Abreviada (WASI). Os resultados indicaram que os pais de crianças com TEA apresentaram maior dificuldade em linguagem pragmática do BAPQ, ainda que não tenham fechado critério para FAA, e nas habilidades comunicativo-pragmáticas de discurso conversacional, prosódia emocional, interpretação de metáforas e de atos de fala indiretos. Foi também verificado que o grupo clínico apresentou menor escore em qi verbal em relação ao grupo comparativo. Conclui-se que maiores dificuldades em habilidades cognitivo-pragmáticas podem ser um aspecto central e provavelmente específico do FAA em familiares de crianças com TEA, assim como dificuldades em aspectos estruturais da linguagem, como inteligência verbal. Estes resultados têm implicações para a compreensão da possível influência de fatores sociocognitivos familiares na gênese do autismo. The broader autism phenotype (BAP) has been defined as subclinical traits that configure the phenotypic expression of genetic susceptibility to the development of Autistic Spectrum Disorder (ASD). The literature has pointed to the relationship between communicative-pragmatic skills and autism. However, there is controversy as to whether these skills constitute a specific feature of the BAP. Still, research with first-degree relatives of people with ASD has not yet been successful in establishing definitive markers. Thus, this study compared the performance of parents of children with (n=26) and without ASD (n=26) in communicative-pragmatic skills tasks and BAP indicators. The Broader Autism Phenotype Questionnaire (BAPQ) and subtests of the Montreal Communication Evaluation Battery (mac Battery), and the Wechsler Abbreviated Intelligence Scale (WASI) were applied. The results indicated that parents of children with ASD had greater difficulty in the pragmatic language of the BAPQ, although they did not meet the criteria for BAP and in communicative-pragmatic skills of conversational speech, emotional prosody, metaphor interpretation, and indirect speech acts. The clinical group also presented a lower verbal iq score compared to the nonclinical group. It is concluded that greater difficulties in cognitive-pragmatic skills may be a central and probably specific aspect of the BAP in family members of children with ASD, as well as difficulties in structural aspects of language, such as verbal intelligence. These results have implications for understanding the possible influence of familiar socio-cognitive factors on the genesis of autism. El fenotipo ampliado del autismo (FAA) se ha definido como trazos subclínicos que configuran la expresión fenotípica de una susceptibilidad genética para el desarrollo del Trastorno del Espectro Autista (TEA). La literatura ha apuntado hacia la relación entre habilidades comunicativo-pragmáticas y autismo. Sin embargo, hay controversia en cuanto a la posibilidad de que esas habilidades constituyan una especificidad del FAA. Así mismo, investigaciones con parientes de primer grado de personas con TEA han fallado en establecer marcadores definitivos. Este estudio comparó el desempeño de padres y madres de niños con TEA (n=26) y sin TEA (n=26) en tareas de habilidades comunicativo-pragmáticas e indicadores de FAA. Fueron aplicados el Broader Autism Phenotype Questionnaire (BAPQ) y subpruebas de la Batería Montreal de evaluación de la Comunicación (Batería MAC), además de la Escala de Inteligencia Wechsler Abreviada (WASI). Los resultados indicaron que los padres de niños con TEA presentaron mayor dificultad en lenguaje pragmático del BAPQ, aunque no hayan cumplido el criterio para FAA, y en las habilidades comunicativo-pragmáticas de discurso conversacional, prosodia emocional, interpretación de metáforas y de actos de habla indirectos. También se verificó que el grupo clínico presentó menor puntaje en ci verbal en relación con el grupo comparativo. Se concluyó que mayores dificultades en habilidades cognitivo-pragmáticas pueden ser un aspecto central y probablemente específico del FAA en familiares de niños con TEA, así como dificultades en aspectos estructurales del lenguaje, como inteligencia verbal. Estos resultados tienen implicaciones para comprender la posible influencia de factores sociocognitivos familiares en la génesis del autismo.

Being able to detect topics and speaker stances in conversations is a key requirement for developing spoken language understanding systems that are personalized and adaptive. In this work, we explore how topic-oriented speaker stance is expressed in conversational speech. To do this, we present a new set of topic and stance annotations of the CallHome corpus of spontaneous dialogues. Specifically, we focus on six stances—positivity, certainty, surprise, amusement, interest, and comfort—which are useful for characterizing important aspects of a conversation, such as whether a conversation is going well or not. Based on this, we investigate the use of neural network models for automatically detecting speaker stance from speech in multi-turn, multi-speaker contexts. In particular, we examine how performance changes depending on how input feature representations are constructed and how this is related to dialogue structure. Our experiments show that incorporating both lexical and acoustic features is beneficial for stance detection. However, we observe variation in whether using hierarchical models for encoding lexical and acoustic information improves performance, suggesting that some aspects of speaker stance are expressed more locally than others. Overall, our findings highlight the importance of modelling interaction dynamics and non-lexical content for stance detection.

Synthesising spontaneous speech is a difficult task due to disfluencies, high variability and syntactic conventions different from those of written language. Using found data, as opposed to lab-rec ...

The communication phenomenon known as conversational entrainment occurs when dialogue partners align or adapt their behavior to one another while conversing. Associated with rapport, trust, and communicative efficiency, entrainment appears to facilitate conversational success. In this work, we explore how conversational partners entrain or align on articulatory precision or the clarity with which speakers articulate their spoken productions. Articulatory precision also has implications for conversational success as precise articulation can enhance speech understanding and intelligibility. However, in conversational speech, speakers tend to reduce their articulatory precision, preferring low-cost, imprecise speech. Speakers may adapt their articulation and become more precise depending on feedback from their listeners. Given the potential of entrainment, we are interested in how conversational partners adapt or entrain their articulatory precision to one another. We explore this phenomenon in 57 task-based dialogues. Controlling for the influence of speaking rate, we find that speakers entrain on articulatory precision, with significant alignment on articulation of consonants. We discuss the potential applications that speaker alignment on precision might have for modeling conversation and implementing strategies for enhancing communicative success in human-human and human-computer interactions.

We present an end-to-end speech recognition model that learns interaction between two speakers based on the turn-changing information. Unlike conventional speech recognition models, our model exploits two speakers' history of conversational-context information that spans across multiple turns within an end-to-end framework. Specifically, we propose a speaker-specific cross-attention mechanism that can look at the output of the other speaker side as well as the one of the current speaker for better at recognizing long conversations. We evaluated the models on the Switchboard conversational speech corpus and show that our model outperforms standard end-to-end speech recognition models., Interspeech 2019

With the growing popularity of social networks, cloud services and online applications, people are becoming concerned about the way companies store their data and the ways in which the data can be applied. Privacy with devices and services operated by the voice are of particular interest. To enable studies in privacy, this paper presents a database which quantifies the experience of privacy users have in spoken communication. We focus on the effect of the acoustic environment on that perception of privacy. Speech signals are recorded in scenarios simulating real-life situations, where the acoustic environment has an effect on the experience of privacy. The acoustic data is complemented with measures of the speakers’ experience of privacy, recorded using a questionnaire. The presented corpus enables studies in how acoustic environments affect peoples’ experience of privacy, which in turn, can be used to develop speech operated applications which are respectful of their right to privacy.

BackgroundSimulated learning environments are increasingly common in interprofessional education (IPE). While reflection is key to simulated learning, little is known about the nature of these conversations during simulation. The aim of this exploratory paper was to quantify communicative features of conversations during interprofessional simulation scenarios between dietetics students, speech-language therapy students and their educators.MethodsConversations between students and educators during the pauses between simulated scenario phases were recorded and transcribed. Student and educator utterances were quantitatively analysed for speech acts, question types and elements of IPE (clinical reasoning, roles and responsibilities, client and family centred care, interprofessional collaboration, clinical procedural tasks).ResultsAcross 1340 utterances from six scenarios, analyses of conversational speech acts and question types highlighted similar patterns of usage between two educators despite different clinical scenarios and professional backgrounds. Educators used a minimally higher proportion of open compared with closed questions, and higher-level problem-solving questions predominated in comparison to simple factual questioning. Educators used more requests for action and attention and students displayed more performative and responsive acts (pConclusionsConversations during pauses in immersive simulated scenarios between educators and students enable rich IPE opportunities and higher-level problem-solving. Educators encouraged students to problem solve within and across disciplines with open questions. Educators provided few factual responses to questions themselves rather diverting questions back to the students. This approach to the analysis of conversation can support educators to evaluate their own communication during interprofessional simulations.

Persistence is the tendency of speakers to repeat the choice of sociolinguistic variant they have recently made in conversational speech. A longstanding debate is whether this tendency toward repetitiveness reflects the direct influence of one outcome on the next instance of the variable, which I call sequential dependence, or the shared influence of shifting contextual factors on proximal instances of the variable, which I call baseline deflection. I propose that these distinct types of clustering make different predictions for sequences of variable observations that are longer than the typical prime-target pairs of typical corpus persistence studies. In corpus ING data from conversational speech, I show that there are two effects to be accounted for: an effect of how many times the /ing/ variant occurs in the 2, 3, or 4-token sequence prior to the target (regardless of order), and an effect of whether the immediately prior (1-back) token was /ing/. I then build a series of simulations involving Bernoulli trials at sequences of different probabilities that incorporate either a sequential dependence mechanism, a baseline deflection mechanism, or both. I argue that the model incorporating both baseline deflection and sequential dependence is best able to produce simulated data that shares the relevant properties of the corpus data, which is an encouraging outcome because we have independent reasons to expect both baseline deflection and sequential dependence to exist. I conclude that this exploratory analysis of longer sociolinguistic sequences reflects a promising direction for future research on the mechanisms involved in the production of sociolinguistic variation.

This paper investigates the use of subband temporal envelope (STE) features and speed perturbation based data augmentation in end-to-end recognition of distant conversational speech in everyday home environments. STE features track energy peaks in perceptual frequency bands which reflect the resonant properties of the vocal tract. Data augmentation is performed by adding more training data obtained after modifying the speed of the original training data. Experiments show that using STE features and speed perturbation based data augmentation helps improving the performance of end-to-end speech recognition on a challenging corpus which was used for the CHiME 2018 speech separation and recognition challenge. STE features provide up to 2.0% relative word error rate (WER) reduction compared to the conventional log-Mel filter-bank (FBANK) features. Data augmentation is used with both features and provides up to 5.2% relative WER reduction. We propose a simple hypothesis selection method to combine the hypotheses produced by the end-to-end systems using FBANK and STE features. This method additionally provides up to 4.7% relative WER reduction.

How do people hear laughter in expressive, unprompted speechƒ What is the range of expressivity and function of laughter in this speech, and how can laughter inform the recognition of higher-level expressive dimensions in a corpusƒ This paper presents a scalable method for collecting natural human description of laughter, transforming the description to a vector of quantifiable laughter dimensions, and deriving baseline classifiers for the different dimensions of expressive laughter. Then, it explores the impact of leveraging nuances of laughter in the recognition of higher-level, general expressive dimensions, discovered in the same way, such as genuine happiness, sarcasm, nervous reflection, and more. The performance of the low-level laughter classifiers is presented, along with the performance of the high-level laughter-aware and laughter-unaware classifiers.

This paper explores contexts associated with errors in transcrip-tion of spontaneous speech, shedding light on human perceptionof disfluencies and other conversational speech phenomena. Anew version of the Switchboard corpus is provided with disfluency annotations for careful speech transcripts, together with results showing the impact of transcription errors on evaluation of automatic disfluency detection., Submitted to INTERSPEECH 2019

Precise timing is crucial for many behaviors ranging from street crossing, conversational speech, to athletic performance. The precision of motor timing has been suggested to result from the strength of phase-amplitude coupling (PAC) between the phase of alpha oscillations (α, 8-12 Hz) and the power of beta activity (β, 14-30 Hz), herein referred to as α-β PAC. The amplitude of β oscillations has been proposed to code for temporally relevant information, and the locking of β power to the phase of α oscillations to maintain timing precision. Motor timing precision has at least two sources of variability: variability of timekeeping mechanism and variability of motor control. There is ambiguity to with of these two factors α-β PAC could be ascribed to. Whether α-β PAC indexes precision of internal timekeeping mechanisms like a stopwatch, or α-β PAC indexes motor control precision is unclear. To disentangle these two hypotheses, we tested how oscillatory coupling at different stages of time reproduction related to temporal precision. Human participants perceived, and subsequently reproduced, a time interval while magnetoencephalography was recorded. The data show a robust α-β PAC during both the encoding and the reproduction of a temporal interval, a pattern which could not be predicted for by the motor control account. Specifically, we found that timing precision resulted from the tradeoff between the strength of α-β PAC during the encoding and during the reproduction of intervals. We interpret these results as supporting evidence for the hypothesis that α-β PAC codes for precision of temporal representations in the human brain.Highlights-Encoding and reproducing temporal intervals implicate α-β PAC.-α-β PAC does not represent solely motor control.-α-β PAC maintains the precision of temporal representations.

This article offers a reading of Kole Omotoso’s The Edifice in which he exposes the themes of insensitivity, callousness, aggressiveness, oppression, marginalization and racial discrimination of the whites against the black people successfully. It is clear that the most important and highly unique technique that the writer uses to advantage and which projects the theme of the novel forward is the one of conversational speech which contributes highly to the success of the novel. This explains the reason the present writer is picking the novel out of all Omotoso’s novels because it makes it more prone to linguistic analysis more than most African literary works. Using linguistic stylistic analytical tool of narrative voice structure and focusing particularly on free direct speech and free direct thought, this paper examines the themes of racial discrimination and oppression of the whites against the black people in the novel. From the analysis, the paper concludes that the implication of this racial discrimination is that all the expressions used in the novel against the black people by the whites are exaggerated, hyperbolic and highly calculated to depict African race in bad light.

espanolSe presentan los resultados iniciales del analisis del fenomeno del habla ascendente registrada en Chiloe. Se trata de enunciados declarativos finales en foco amplio que manifiestan un aumento en la frecuencia fundamental, a pesar de que en la mayoria de las variedades reportadas del espanol se producen con un descenso paulatino del tono. Se analizan los primeros datos acusticos de muestras de conversacion y se hacen las observaciones fonologicas correspondientes. Se presenta evidencia acustica del ascenso, una propuesta de etiquetaje fonologico y se plantean interrogantes y posibilidades nuevas de investigacion EnglishThis research paper presents initial findings regarding the analysis of rising intonation contours in final, broad focus declarative utterances recorded in Chiloe (Chile). These utterances—unlike the great majority that have been reported in different varieties of Spanish and characterized by a falling nuclear configuration—exhibit a rising terminal pitch. The first acoustic data of conversational speech are analysed and preliminary phonological observations are made. Acoustic evidence of the rise is included, together with tentative phonological labelling and, finally, questions are raised and possibilities for further investigation are proposed.

Objective: This is the first study to investigate the behavioral nature (topography) of stuttering in Cantonese. Cantonese, a Sino-Tibetan language, is both tonal and syllable-timed. Previous studies of stuttering topography have mainly been in Western languages, which are mainly stress-timed. Methods: Conversational speech samples were collected from 24 native Cantonese-speaking adults who stuttered. Six consecutive stuttering moments from each participant were analyzed using the Lidcombe behavioral data language (LBDL). A complexity analysis based on the LBDL was developed to indicate the proportion of multiple-behavior stuttering moments for each participant. Results: There was no significant difference in the frequency of the 7 LBDL behaviors. Almost half the stuttering moments across participants were reported as complex, containing more than 1 stuttering behavior, and stuttering complexity correlated significantly with stuttering severity. Conclusions: These preliminary findings require replication because of their important theoretical and clinical implications. Differences in topography across languages have the potential to contribute to our understanding of the nature of stuttering. Clinically, the recognition of such differences may assist practitioners in identifying stuttering, for example when screening for early stuttering. The LBDL complexity score developed in this study has the potential to be used in other languages.

OBJECTIVEPatients with Parkinson disease (PD) who undergo subthalamic nucleus (STN) deep brain stimulation (DBS) often develop a deterioration in speech performance, but there is no clear consensus on the specific effects seen or the mechanism involved and little description of the impact of DBS on conversational speech. Furthermore, there has been no fiber tract connectivity analysis to identify the structures potentially modulated by DBS to cause such deficits. The main objective of this study was to quantify spontaneous speech performance and identify potential involvement of the dentatorubrothalamic tract (DRTt) in patients who underwent STN DBS, because this tract has been implicated in speech deterioration.METHODSSpontaneous speech samples were obtained with STN DBS in both on and off modes in 35 patients with PD and assessed across multiple domains. Diffusion tensor imaging tractography seeded from the therapeutic DBS contacts was performed to identify the fiber tracts involved and, specifically, the DRTt. The position of active electrode contacts was assessed relative to that of the STN.RESULTSFifteen patients with akinetic-rigid (AR) PD and 20 with tremor-dominant (TD) PD subtypes were identified. In the AR-PD subgroup of patients, in whom there was DRTt involvement, 71% demonstrated much better overall speech and largely improved or unchanged fluency in the DBS-off condition. In patients with TD PD with DRTt involvement, 50% demonstrated better overall speech in the off condition, and equivocal results regarding improved or worsened fluency were found. When there was minimal DRTt involvement, 75% of patients with AR PD had better overall speech in the DBS-on condition and better or minimal fluency changes. Similarly, 83% of patients with TD PD with minimal DRTt involvement had better or minimal overall speech and fluency changes in the on condition. More medially placed left electrode contacts were associated with more DRTt involvement in 77% of patients (10 of 13).CONCLUSIONSTo the authors' knowledge, this is the first study to have investigated a specific fiber tract involved in STN DBS in different subtypes of PD relative to its impact on spontaneous speech. At optimal therapeutic programming of STN DBS, overall spontaneous speech and fluency were affected more negatively in patients with AR PD than in those with TD PD when there was DRTt involvement. After fiber tract analysis and modeling, it was found that medially positioned left electrode contacts more often involved fibers of the DRTt. If possible, avoidance of the DRTt by using active electrode contacts that are positioned less medially, specifically in patients with AR PD, might result in less speech deterioration.

Motivated by the quasi-categorical reduced forms of disyllabic words produced in Chinese conversational speech, a frequency-based selection procedure of typical pronunciation by disyllabic word type and reduction degree is proposed in this paper. This variant-selection algorithm utilizes techniques of free phone recognition and phonetic similarity score calculation to account for Chinese syllable structure. Four reduction types are suggested by considering the presence of a within-word syllable boundary: Citation form-like reduction, marginal segment deletion, nuclei merger, and syllable merger. The results show that the most frequent reduction types for disyllabic words in Chinese conversation are citation form-like reduction and syllable merger. In particular, high-frequency disyllabic words preferentially take the extreme syllable-merger form. As shown in the analysis, segmental reduction in Chinese disyllabic words is morphology-dependent. It is also related to the prosodic position at which a disyllabic word is produced as well as the temporal quality of the word. Finally, in the automatic speech recognition experiments, the performance was improved by adding a small number of variants selected by the algorithm to the pronunciation dictionary of the system.

In recent years, the interest in research in speech understanding and spoken interaction has soared due to the emergence of virtual personal assistants. However, while the ability of these agents to recognise conversational speech is maturing rapidly, their ability to understand and interact is still limited. At the same time we have witnessed the development of the number of models based on machine learning that made a huge impact on spoken language understanding accuracies and the interaction quality overall. This special issue brings together a number of articles that tackle different aspects of spoken language understanding and interaction: clarifications in dialogues, adaptation to different domains, semantic tagging and error handling. These studies all have a common purpose of building human-like conversational systems.

Background A considerable amount of research has identified inhibition differences, including slower inhibition of manual responses, in people who stutter. Recent investigations have failed to link slowed motor inhibition with overt stuttering severity. This study investigated the potential relationship between slowed manual response inhibition and the negative impact of stuttering upon individual lives of adults who stutter (AWS). Methods Thirty-four adults (AWS, n = 17; AWNS, n = 17) matched by nonverbal IQ completed a manual stop-signal task and provided a conversational speech sample. Motor inhibition latency for AWS and AWNS were compared. For AWS, motor inhibition latency was compared to the four subsections of the Overall Assessment of the Speaker’s Experience with Stuttering (OASES; Yaruss & Quesal, 2006; General Information, Reactions to Stuttering, Communication in Daily Situations, Quality of Life). Results Similar to previous studies, AWS were significantly slower to inhibit inaccurate manual responses than AWNS. Quality of Life subtest of the OASES was found to significantly predict inhibition latency. Conclusion These data replicate findings that indicate that AWS exhibit slower manual inhibition latency, and suggest that these inhibition differences may be associated with an individual’s negative experience with stuttering rather than stuttering severity.

Background: Augmented reality (AR) has great potential for improving image-guided neurosurgical procedures, but until recently, hardware was mostly custom-made and difficult to distribute. Currently, commercially available low-cost AR devices offer great potential for neurosurgery, but reports on technical feasibility are lacking. The goal of this pilot study is to evaluate the feasibility of using a low-cost commercially available head-mounted holographic AR device (the Microsoft Hololens) in the operating room. The Hololens is operated by performing specific hand gestures, which are recognized by the built-in camera of the device. This would allow the neurosurgeon to control the device "touch free" even while wearing a sterile surgical outfit.Methods: The Hololens was tested in an operating room under two lighting conditions (general background theatre lighting only; and general background theatre lighting and operating lights) and wearing different surgical gloves (both bright and dark). All required hand gestures were performed, and voice recognition was evaluated against background noise consisting of two nurses talking at conversational speech level.Results: Wearing comfort was sufficient, with and without regular glasses. All gestures were correctly classified regardless of lighting conditions or the sort of sterile gloves. Voice recognition was good. The visibility of the holograms was good if the device was configured to use high brightness for display.Conclusions: We demonstrate that using a commercially available low-cost head-mounted holographic AR device is feasible in a sterile surgical setting, under different lighting conditions and using different surgical gloves. Given the availability of freely available software for application development, neurosurgery can benefit from new opportunities for image-guided surgery.

Language model is a vital component in modern automatic speech recognition (ASR) systems. Since “one-size-fits-all” language model works suboptimally for conversational speeches, language model adaptation (LMA) is considered as a promising solution for solving this problem. In order to compare the state-of-the-art LMA techniques and systematically demonstrate their effect in conversational speech recognition, we develop a novel toolkit named Chameleon, which includes the state-of-the-art cache-based and topic-based LMA techniques. This demonstration does not only vividly visualize underlying working mechanisms of a variety of the state-of-the-art LMA models but also provide an interface for the user to customize the hyperparameters of them. With this demonstration, the audience can experience the effect of LMA in an interactive and real-time fashion. We wish this demonstration would inspire more research on better language model techniques for ASR.

This study examines cross-generational transmission of a sociophonetic variable in a heritage language context. Voiceless stop aspiration is a sociophonetic variable in Calabrian Italian, indexing socio-cultural values about the speaker’s social and geographical origin. We investigate the production of voiceless stops by three generations of Calabrian Italians (immigrants and the next two generations) in Toronto, via acoustic and auditory analysis of nearly 5000 tokens from conversational speech in Calabrian Italian. Both Italian and English use long-lag VOT, but they differ in its phonological distribution: long-lag VOT is preferentially associated with pre-tonic, word-initial stops in English and with post-tonic, post-sonorant or geminate stops in Calabrian Italian. We show that, in heritage Calabrian Italian in Toronto, both phonetic implementation (cued by VOT duration) and phonological distribution of aspiration (as cued by perceived aspiration rate across phonological contexts) change cross-generationally, but some changes are non-linear, as third generation speakers appear to reproduce some patterns attested in the speech of first generation speakers. External variables such as the sex of the speakers modulate the cross-generational effects, with males producing more aspirated stops and exhibiting a more conservative behavior in certain phonetic contexts.

В статье исследуются конкретные слова: устаревший термин лихоимство и термин взятка. Значение этих терминов прослеживается в их историческом развитии. При этом используются сведения из законов, словарей, других документов. Особое внимание уделяется современным терминологическим аналогам, отражающим российское законодательство. Исследуемые термины анализируются с точки зрения значения, в котором они употребляются в текстах документов и разговорной речи, с позиций их статуса в русском литературном языке, а также с точки зрения стилистической окраски и сферы употребления на протяжении столетий., The article analyzes specific words: the outdated term lihoimstvo and the term vzjatka (both mean bribe). The meaning of these terms is traced in their historical development. The information from laws, dictionaries and other documents is used. Particular attention is paid to modern terminological analogues reflecting Russian legislation. The selected terms are analyzed from the point of view of the meaning in which they are used in the texts of documents and conversational speech, from the standpoint of their status in the Russian literary language, as well as from the point of view of stylistic coloring and sphere of use over the centuries.

Although the way tones are acquired by second or foreign language learners has attracted some scholarly attention, detailed knowledge of the factors that promote efficient learning is lacking. In this article, we look at the effect of visual cues (comparing audio-only with audio-visual presentations) and speaking style (comparing a natural speaking style with a teaching speaking style) on the perception of Mandarin tones by non-native listeners, looking both at the relative strength of these two factors and their possible interactions. Both the accuracy and reaction time of the listeners were measured in a task of tone identification. Results showed that participants in the audio-visual condition distinguished tones more accurately than participants in the audio-only condition. Interestingly, this varied as a function of speaking style, but only for stimuli from specific speakers. Additionally, some tones (notably tone 3) were recognized more quickly and accurately than others.

Human verbal communication is a complex phenomenon involving dynamics that normally result in the alignment of participants on several modalities, and across various linguistic domains. We examined here whether such dynamics occur also for paralinguistic events, in particular, in the case of laughter. Using a conversational corpus containing dyadic interactions in three languages (French, German and Mandarin Chinese), we investigated three measures of alignment: convergence, synchrony and agreement. Support for convergence and synchrony was found in all three languages, although the level of support varied with the language, while the agreement in laughter type was found to be significant for the German data. The implications of these findings towards a better understanding of the role of laughter in human communication are discussed.

Prosodic cues in conversational speech aid listeners in discerning a message. We investigate whether acoustic cues in spoken dialogue can be used to identify the importance of individual words to the meaning of a conversation turn. Individuals who are Deaf and Hard of Hearing often rely on real-time captions in live meetings. Word error rate, a traditional metric for evaluating automatic speech recognition, fails to capture that some words are more important for a system to transcribe correctly than others. We present and evaluate neural architectures that use acoustic features for 3-class word importance prediction. Our model performs competitively against state-of-the-art text-based word-importance prediction models, and it demonstrates particular benefits when operating on imperfect ASR output., Comment: 8 pages, 2 figures

Contains fulltext : 214126.pdf (Publisher’s version ) (Open Access) Interspeech 2019, 15 september 2019

Euskaraz hitz-bukaerako herskariek oso maiztasun txikia dute hiztegian, baina oso maiztasun handia solasean; /-t/ eta /-k/ kontsonanteak maiztasun handiko atzizki batzuetan aurkitzen baitira. Lan honetan hitz-bukaerako herskarien gauzatze fonetikoa aztertzen dugu, eta bereziki bi hitzen arteko kontsonante taldeetan jartzen dugu arreta. Datuak elkarrizketa libretik atera ditugu. Intentsitate neurketak erakusten du talde horietan kontsonanteak oso ahulak direla gehienetan. Espektrogramak miatuz baieztatzen dugu /-k/ eta /-t/ kontsonanteak beste kontsonante baten aurrean gehiengoetan galtzen edo hurbilkari ahostun gisa ahoskatzen direla. Bestalde, ahoskabetze gertakariak (adib. /k#b/ --> [p]) ez dira batere normalak gure corpusean.

International audience; In order to model hesitations for technical applications such as conversational speech synthesis, it is desirable to understand interactions between individual hesitation markers. In this study, we explore a pair of markers that has been subject to many discussions: silences and fillers. While it is generally acknowledged that fillers occur in two distinct forms, um and uh, it is not agreed on whether these forms systematically influence the form of associated silences. This notion will be investigated on a small dataset of English spontaneous speech data and the measure of distance between filler and silence will be introduced to the analyses. Results suggest that filler type influences associated silence duration systematically and that silences tend to gravitate towards fillers in utterances, exhibiting systematically lower duration when preceding them. These results provide valuable insights for improving existing hesitation models.

The study is aimed to extract discourse relations patterns in conversational speech of subjects with Alzheimer’s Disease (AD) and adults with healthy aging processes using the Rhetorical Structure Theory (RST). By means of the RST, we analyzed semi-structured interviews of native Spanish speakers. Seven subjects were in the mild, moderate or advanced stages of AD, and 6 were cognitively intact individuals. The procedure involved the segmentation of each conversational discourse into Semantic Dialog Units (SDUs), the labeling of their rhetorical relations and the construction of tree diagrams. We perform acorrelation analysis to determine the significance of the use of rhetorical relations for each group. We found a significantly (p-value < .05) lower rhetorical relations production density in subjects with AD. We also observed that most rhetorical relations used by healthy older subjects were Elaboration, Concession, Interpretation, Non-Volitional Cause, Solutionhood and Volitional Result.

International audience; Interactions between speech style and coarticulation are investigated by examining local, non-local, anticipatory and carryover contextual effects on vowels in two French corpora of conversational and journalistic speech. C-to-V coarticulation is analyzed on 22 k tokens of /i, E, a, u, ɔ/(/E/=/e, ɛ/) 50-to-80 ms long. Contextual effects are measured as F2 changes in relation to the adjacent consonant (alveolar vs. uvular) in CV1 and V1C sequences. V-to-V coarticula-tion is analyzed on 33 k V1C(C)V2 sequences with V1 = /e, ɛ, o, ɔ, a/ falling within the same range of duration, and V2 either high/mid-high or low/mid-low. Contextual effects are measured as F1 changes as function of V2 height. Results show more local C-to-V coarticulation in conversational than in journalistic speech, as previously found for other languages. Interestingly, this interaction is clearer for all vowels in V1C, whereas coarticulation in CV1 is affected by style for non-high vowels only. V-to-V coarticulation is also found in both corpora but is modulated by style only for mid-front vowels and in the opposite direction (i.e. more overlap in journalistic than in conversational speech). Findings are interpreted in light of dynamic models of speech production and of a phono-logical account of French V-to-V harmony.

An ASR system usually does not predict any punctuation or capitalization. Lack of punctuation causes problems in result presentation and confuses both the human reader andoff-the-shelf natural language processing algorithms. To overcome these limitations, we train two variants of Deep Neural Network (DNN) sequence labelling models - a Bidirectional Long Short-Term Memory (BLSTM) and a Convolutional Neural Network (CNN), to predict the punctuation. The models are trained on the Fisher corpus which includes punctuation annotation. In our experiments, we combine time-aligned and punctuated Fisher corpus transcripts using a sequence alignment algorithm. The neural networks are trained on Common Web Crawl GloVe embedding of the words in Fisher transcripts aligned with conversation side indicators and word time infomation. The CNNs yield a better precision and BLSTMs tend to have better recall. While BLSTMs make fewer mistakes overall, the punctuation predicted by the CNN is more accurate - especially in the case of question marks. Our results constitute significant evidence that the distribution of words in time, as well as pre-trained embeddings, can be useful in the punctuation prediction task., Accepted for Interspeech 2018 Conference

International audience; The CHiME challenge series aims to advance robust automatic speech recognition (ASR) technology by promoting research at the interface of speech and language processing, signal processing , and machine learning. This paper introduces the 5th CHiME Challenge, which considers the task of distant multi-microphone conversational ASR in real home environments. Speech material was elicited using a dinner party scenario with efforts taken to capture data that is representative of natural conversational speech and recorded by 6 Kinect microphone arrays and 4 binaural microphone pairs. The challenge features a single-array track and a multiple-array track and, for each track, distinct rankings will be produced for systems focusing on robustness with respect to distant-microphone capture vs. systems attempting to address all aspects of the task including conversational language modeling. We discuss the rationale for the challenge and provide a detailed description of the data collection procedure, the task, and the baseline systems for array synchronization, speech enhancement, and conventional and end-to-end ASR.

In English, nasal place assimilation occurs across word boundaries, such as 'ten bucks 'pronounced as 'te[m] bucks'. Assimilation can be viewed as a reduction or loss of the assimilation target’s place cue (/n/ in 'ten'), and simultaneously as an enhancement of the assimilation trigger’s place cue (/b/ in 'bucks') by spreading its place cue earlier in the signal. A message-oriented phonological approach predicts that assimilation is sensitive to the relative contextual inferrability of both the target and trigger words: More assimilation should be observed for more contextually predictable target words, while 'less 'assimilation should be observed for contextually more predictable 'trigger 'words. These predictions deviate from accounts that view assimilation solely as reduction. To test these predictions, sequences which license assimilation were extracted from a conversational speech corpus. Both categorical assimilation (based on close phonetic transcription) and gradient acoustic assimilation (based on F2) were analyzed. As predicted, assimilation was more likely both when a target like 'ten 'was high in predictability and when its trigger 'bucks 'was low in predictability. Assimilation thus serves as both reduction and enhancement, and can be used to manage redundancy in the speech signal. More broadly, this constitutes evidence for the influence of communicative pressures on phonology.

Formulaic expressions naturally convey affective content. The unique formal and functional characteristics of idioms, slang, expletives, proverbs, conversational speech formulas, and the many other conventional expressions in this repertory have been well-described: these include unitary form, conventionalized and non-literal meanings, and reliance on social context. Less highlighted, but potent, is the intrinsic presence of affective meaning. Expletives, for example, signal strong emotion. Idioms, too, inherently communicate emotional connotations, and conversational speech formulas allow for empathetic bonding and humor. The built-in affective content of formulaic expressions, in combination with their other unique characteristics, is compatible with the proposal that brain structures other than those representing grammatical language are in play in producing formulaic expressions. Evidence is presented for a dual process model of language, whereby a right hemisphere-subcortical system modulates formulaic language.

Bipolar Disorder is a chronic psychiatric illness characterized by pathological mood swings associated with severe disruptions in emotion regulation. Clinical monitoring of mood is key to the care of these dynamic and incapacitating mood states. Frequent and detailed monitoring improves clinical sensitivity to detect mood state changes, but typically requires costly and limited resources. Speech characteristics change during both depressed and manic states, suggesting automatic methods applied to the speech signal can be effectively used to monitor mood state changes. However, speech is modulated by many factors, which renders mood state prediction challenging. We hypothesize that emotion can be used as an intermediary step to improve mood state prediction. This paper presents critical steps in developing this pipeline, including (1) a new in the wild emotion dataset, the PRIORI Emotion Dataset, collected from everyday smartphone conversational speech recordings, (2) activation/valence emotion recognition baselines on this dataset (PCC of 0.71 and 0.41, respectively), and (3) significant correlation between predicted emotion and mood state for individuals with bipolar disorder. This provides evidence and a working baseline for the use of emotion as a meta-feature for mood state monitoring., Interspeech 2018

Millions of homes worldwide enjoy access to digital content and services through smart speakers such as Amazon's Echo and Google's Home. Promotional materials and users' own videos typically show homes that have many well-resourced rooms, with good power and data infrastructures. Over the last several years, we have been working with slum communities in India, whose dwellings are usually very compact (one or two rooms), personal home WiFi is almost unheard of, power infrastructures are far less robust, and financial resources put such smart speakers out of individual household reach. Inspired by the "hole in the wall" internet-kiosk programme, we carried out workshops with slum inhabitants to uncover issues and opportunities for providing a smart-speaker-type device in public areas and passageways. We designed and deployed a simple probe that allowed passers-by to ask and receive answers to questions. In this paper, we present the findings of this work, and a design space for such devices in these settings.

Listening effort increases when perceiving speech masked by noise and when listeners are non-native speakers of the target language (Zekveld and Kramer, 2014; Borghini and Hazan, 2018). Listener-oriented clear speech (CS) improves word recognition in noise and recognition memory (Keerstock and Smiljanic, 2018). It is unclear, however, whether the CS benefit can be explained by reduced listening effort during processing of easier-to-understand speech. The goal of this study was to examine the effect of CS on listening effort associated with degraded listening conditions differing in the source of difficulty and underlying cognitive mechanism: energetic masking, informational masking, and non-native listening. Native and non-native English listeners performed a word recognition in noise task and a visual response task, i.e., pressing keys to indicate the orientation of an arrow on the screen, first separately and then simultaneously. In the word recognition task, listeners repeated back sentences produced in CS and conversational speech and mixed with speech-shaped noise and 2-talker babble. The presentation will discuss analyses aimed at uncovering how hyper-articulated CS and masker type influence listening effort for native and non-native listeners as measured by objective and subjective dual-task cost.

RésuméLa distribution des variables utilisées pour l'expression de l'ultériorité a fait l'objet de nombreuses études, centrées sur l'oral conversationnel ou sur l’écrit. Cet article apporte un éclairage inédit sur la question en considérant les marqueurs du futur dans un contexte communicatif d'oral préparé (le bulletin météo télévisé), et dans une optique diatopique (comparaison d'un corpus français et d'un corpus québécois). L'analyse distributionnelle indique une distribution des variables spécifique au contexte discursif. Par ailleurs, l'analyse multivariée par Goldvarb X révèle une variation diatopique et l'influence de certains facteurs linguistiques, en particulier celui du type de verbe, de même que des contraintes particulières pour les deux communautés étudiées.

PurposeThe production of formulaic expressions (conversational speech formulas, pause fillers, idioms, and other fixed expressions) is excessive in the left hemisphere and deficient in the right hemisphere and in subcortical stroke. Speakers with Alzheimer's disease (AD), having functional basal ganglia, reveal abnormally high proportions of formulaic language. Persons with Parkinson's disease (PD), having dysfunctional basal ganglia, were predicted to show impoverished formulaic expressions in contrast to speakers with AD. This study compared participants with PD, participants with AD, and healthy control (HC) participants on protocols probing production and comprehension of formulaic expressions.MethodSpontaneous speech samples were recorded from 16 individuals with PD, 12 individuals with AD, and 18 HC speakers. Structured tests were then administered as probes of comprehension.ResultsThe PD group had lower proportions of formulaic expressions compared with the AD and HC groups. Comprehension testing yielded opposite contrasts: participants with PD showed significantly higher performance compared with participants with AD and did not differ from HC participants.ConclusionsThe finding that PD produced lower proportions of formulaic expressions compared with AD and HC supports the view that subcortical nuclei modulate the production of formulaic expressions. Contrasting results on formal testing of comprehension, whereby participants with AD performed significantly worse than participants with PD and HC participants, indicate differential effects on procedural and declarative knowledge associated with these neurological conditions.