Speech Model Research Articles

УКРАЇНОЦЕНТРИЗМ І ПРОПАГАНДА: РЕАЛІЇ РЕГІОНАЛЬНОГО ТЕЛЕМОВЛЕННЯ ЛУГАНЩИНИ

The article considers the process of development of Luhansk regional television since the beginning of the military aggression of the Russian Federation against Ukraine in 2014. This period is characterized by physical occupation of Luhansk and Donetsk regions, annexation of the Republic of Crimea, aggressive hostilities under the guise of illegal armed groups. , economic and advocacy activities. The analysis of the functioning of the broadcasting system in Luhansk region showed that the transformation of the regional information field and mass information activities were significantly influenced by the accompanying military aggression - hybrid and information war. Russia's aggression against Ukraine has changed the constant mass communication activities of local, regional and national media. In May 2014, TV companies in Luhansk and Donetsk regions were seized by pro-Russian militants and later "nationalized" by the terrorist groups «LNR» and «DNR». The physical occupation of part of the territory of Luhansk (ORLO - occupied districts of Luhansk region) and Donetsk (occupied districts of Donetsk region (ORDO)) regions led to the rapid annexation of part of the information space to saturate it with Russian propaganda and local separatist television channels. At the same time, a Russian propaganda model of speech was introduced in ORDLO. Instead, the revived in evacuation channels of Ukrainian regional television (IRTA, LOT (and its successor UA:Donbass) improved the technical base of TV studios, used the latest technologies, changed the quality of content, which became intelligent, diverse, focused, including audience territories and is based on Ukraine-centric positions.

Quasi-Periodic Parallel WaveGAN: A Non-Autoregressive Raw Waveform Generative Model With Pitch-Dependent Dilated Convolution Neural Network

In this paper, we propose a quasi-periodic parallel WaveGAN (QPPWG) waveform generative model, which applies a quasi-periodic (QP) structure to a parallel WaveGAN (PWG) model using pitch-dependent dilated convolution networks (PDCNNs). PWG is a small-footprint GAN-based raw waveform generative model, whose generation time is much faster than real time because of its compact model and non-autoregressive (non-AR) and non-causal mechanisms. Although PWG achieves high-fidelity speech generation, the generic and simple network architecture lacks pitch controllability for an unseen auxiliary fundamental frequency ($F_{0}$) feature such as a scaled $F_{0}$. To improve the pitch controllability and speech modeling capability, we apply a QP structure with PDCNNs to PWG, which introduces pitch information to the network by dynamically changing the network architecture corresponding to the auxiliary $F_{0}$ feature. Both objective and subjective experimental results show that QPPWG outperforms PWG when the auxiliary $F_{0}$ feature is scaled. Moreover, analyses of the intermediate outputs of QPPWG also show better tractability and interpretability of QPPWG, which respectively models spectral and excitation-like signals using the cascaded fixed and adaptive blocks of the QP structure.

FSCNet: Feature-Specific Convolution Neural Network for Real-Time Speech Enhancement

In recent years, convolutional neural networks (CNNs) have been widely exploited in deep neural network (DNN)-based speech enhancement methods. However, the representation power of CNNs for speech modeling is limited because of the spatial-agnostic convolution kernels. This letter proposes a novel feature-specific convolution neural network (FSCNet) for real-time speech enhancement. In FSCNet, the encoder and decoder are adopted for forward and inverse feature space transformation, respectively. The denoising module based on the feature-specific convolution (FSC) is employed to enhance the generated deep features. Leveraging the long-term global contexts and considering the importance of each feature channel for speech modeling, the convolution kernels of FSC are dynamically parameterized in each time-frequency location. A function-constrained loss is further proposed to train the FSCNet, ensuring the encoder, denoising modules and decoder can function as expected. Experimental results show that the proposed FSCNet outperforms the state-of-the-art denoising algorithms in terms of five objective evaluation metrics and model size.

ЯЗЫКОВАЯ ЛИЧНОСТЬ РЕКТОРА ВУЗА

This article presents the main approaches to the study of the linguistic personality, considers the speech model of the communicant's behavior, which includes the communicative priorities of the linguistic personality, since the individual's belonging to the linguocultural community determines his communicative behavior.

Cortical processing of distracting speech in noisy auditory scenes depends on perceptual demand

Selective attention is essential for the processing of multi-speaker auditory scenes because they require the perceptual segregation of the relevant speech (“target”) from irrelevant speech (“distractors”). For simple sounds, it has been suggested that the processing of multiple distractor sounds depends on bottom-up factors affecting task performance. However, it remains unclear whether such dependency applies to naturalistic multi-speaker auditory scenes.In this study, we tested the hypothesis that increased perceptual demand (the processing requirement posed by the scene to separate the target speech) reduces the cortical processing of distractor speech thus decreasing their perceptual segregation. Human participants were presented with auditory scenes including three speakers and asked to selectively attend to one speaker while their EEG was acquired. The perceptual demand of this selective listening task was varied by introducing an auditory cue (interaural time differences, ITDs) for segregating the target from the distractor speakers, while acoustic differences between the distractors were matched in ITD and loudness. We obtained a quantitative measure of the cortical segregation of distractor speakers by assessing the difference in how accurately speech-envelope following EEG responses could be predicted by models of averaged distractor speech versus models of individual distractor speech. In agreement with our hypothesis, results show that interaural segregation cues led to improved behavioral word-recognition performance and stronger cortical segregation of the distractor speakers. The neural effect was strongest in the δ-band and at early delays (0 − 200 ms).Our results indicate that during low perceptual demand, the human cortex represents individual distractor speech signals as more segregated. This suggests that, in addition to purely acoustical properties, the cortical processing of distractor speakers depends on factors like perceptual demand.

Poetic Discourse in the Early Works of Sigitas Geda

The article examines poetic discourse in the early works of Sigitas Geda— these include Pėdos [Footprints, 1966], Geda’s first published collection of poems, and Strazdas [A Thrush, 1967], a long narrative poem. Poetic discourse is loosely defined in the article as a kind of modelling of meaningful speech in textual practice. The particular literary works selected are read as manifestations of a type of poetic discourse. The article presents an interpretative explication of this particular type and contrasts this approach with established mythopoetic readings of Geda’s oeuvre. In an attempt to examine poetic discourse enunciation, the article describes qualities of the textual fabric, e.g. modes of cohesion, part-towhole relations, collocations, dynamic vectors (such as crescendos), figures of the speaker, object representation, etc.The article singles out and discusses three aspects of the works in question: the particularities of plural enunciation and the way the enunciator projects itself as the ground for the totality of the represented world; the particular cognitive semiotic construct wherein a traditional landscape is represented as a diagrammatic sign that generates signification related to the past, integrating selected modern objects; a micro-plot of ecstatic experience by way of an immediate subjective encounter with the immensity of the past; and epiphanic images of primitive cosmology that emerge in moments of extreme intensity, as attained in the process of enunciation by the selfprojecting imagination.

Automated assessment of speech production and prediction of MCI in older adults

The population of older adults is growing dramatically and, with it comes increased prevalence of neurological disorders, including Alzheimer’s disease (AD). Though existing cognitive screening tests can aid early detection of cognitive decline, these methods are limited in their sensitivity and require trained administrators. The current study sought to determine whether it is possible to identify persons with mild cognitive impairment (MCI) using automated analysis of spontaneous speech. Participants completed a brief neuropsychological test battery and a spontaneous speech task. MCI was classified using established research criteria, and lexical-semantic features were calculated from spontaneous speech. Logistic regression analyses compared the predictive ability of a commonly-used cognitive screening instrument (the Modified Mini Mental Status Exam, 3MS) and speech indices for MCI classification. Testing against constant-only logistic regression models showed that both the 3MS [χ 2(1) = 6.18, p = .013; AIC = 41.46] and speech indices [χ 2(16) = 32.42, p = .009; AIC = 108.41] were able to predict MCI status. Follow-up testing revealed the full speech model better predicted MCI status than did 3MS (p = .049). In combination, the current findings suggest that spontaneous speech may have value as a potential screening measure for the identification of cognitive deficits, though confirmation is needed in larger, prospective studies.

Some Issues Related To The Study Of Drama Language: The Use Of Dialog

The article describes some features of dialogue introduction as a speech structure in different literary genres. Dialogues in drama and prose are correlated with the dialogue model of colloquial speech.

Ликоугрожающие речевые акты в академической среде (разностатусная коммуникация)

The paper aims to describe productive models of face-threatening speech acts in split-level academic communication. The authors analyse speech acts with different levels of threats to a speaker’s positive face: invitation, request (with different degrees of imposition) and proposal. Special attention is paid to analysing the speech act “advice” functioning in institutional communication. Scientific originality of the study lies in the fact that the researchers analyse the use of face-threatening speech acts in split-level academic communication within the framework of the discursive approach. The research findings are as follows: the basic pragmatic characteristics of speech acts under study are identified.

Second Language Learners’ Attitudes Toward French Varieties: The Roles of Learning Experience and Social Networks

AbstractPeople often believe that some language varieties are more prestigious than others, which can trigger speech‐centered biases and inform social judgments of the speaker. However, it is largely unknown what types of language experience and exposure might mitigate language biases, especially for second language (L2) learners. The goal of this study was to investigate this issue by focusing on L2 French learners’ attitudes toward European and Quebec varieties of French. L2 French learners in Montreal (N = 106) rated 2 audios recorded by native speakers from France in a listening comprehension task, with 1 of the 2 speakers introduced as a speaker of Quebec French. The learners described their language learning experience, filled out a French social network questionnaire, and completed a French proficiency test. Results revealed some evidence of reverse linguistic stereotyping, with learners preferring to speak like one speaker significantly more than the other, based on the speaker's assumed identity, not actual speech. Four of the 6 speaker ratings were also associated with participants’ oral proficiency scores, social network density, and positive experiences in Quebec. Findings have implications for the use of speech models in L2 teaching and for the mitigation of language‐centered biases in L2 classrooms.

Improving Automatic Speech Recognition and Speech Translation via Word Embedding Prediction

In this article, we target speech translation (ST). We propose lightweight approaches that generally improve either ASR or end-to-end ST models. We leverage continuous representations of words, known as word embeddings, to improve ASR in cascaded systems as well as end-to-end ST models. The benefit of using word embedding is that word embedding can be obtained easily by training on pure textual data, which alleviates data scarcity issue. Also, word embedding provides additional contextual information to speech models. We motivate to distill the knowledge from word embedding into speech models. In ASR, we use word embeddings as a regularizer to reduce the WER, and further propose a novel decoding method to fuse the semantic relations among words for further improvement. In the end-to-end ST model, we propose leveraging word embeddings as an intermediate representation to enhance translation performance. Our analysis shows that it is possible to map speech signals to semantic space, which motivates future work on applying the proposed methods in spoken language processing tasks.

Phonetic detail encoding in explaining boundary-modulated coarticulation

This study examines the acoustic outputs in boundary coarticulation and attempts to explain the native and nonnative production differences in relation to speech planning. English consonant clusters across words produced by English (L1) and Mandarin (L2) speakers are analyzed. This study focuses on two segmental features (place of articulation and voicing) and computes three acoustic-based measurements (internal release, closure durations and closure shortening). The results find a lesser coarticulation degree in L2 production, and differing acoustic deviations varying with speakers’ proficiency level. The results also suggest that cross-word planning involving place of articulation is robust in both L1 and L2 production, while cross-word voicing encoding is not available in L2 speech. The asymmetry in L2 encoding is compatible with a competitive gestural selection analysis, and the implications of L1 production have been interpreted as in favor of a richer mental representation and finer-level execution than the syllable-sized programming. The preliminary results support the core assumption that L1 and L2 coarticulation differences arise from phonological and phonetic encoding. This study adds cross-linguistic output-based evidence on speech planning and encourages speech models to further specify the very late stage of phonetic encoding.

Topic Modeling and Characterization of Hate Speech against Immigrants on Twitter around the Emergence of a Far-Right Party in Spain

In this paper, we sought to model and characterize hate speech against immigrants on Twitter in Spain around the appearance of the far-right party Vox. More than 240,000 tweets that included the term ‘Vox’ between November 2018 and April 2019 were automatically collected and analyzed. Only 1% of the sample included hate speech expressions. Within this subsample of 1977 messages, we found offenses (56%), incitements to hate (42%), and violent speech (2%). The most frequent terms used were classified into five categories: Spain, Immigration, Government, Islam, and Insults. The most common features were foul language, false or doubtful information, irony, distasteful expressions, humiliation or contempt, physical or psychological threats, and incitement to violence. Using unsupervised topic modeling, we found that the four underlying topics (control of illegal immigration, economic assistance for immigrants, consequences of illegal immigration, and Spain as an arrival point for African immigrants and Islamist terrorism) were similar to those in the discourse of Vox. We conclude that the hate speech against immigrants produced around Vox, and not necessarily by Vox, followed the general patterns of this type of speech detected in previous works, including Islamophobia, offensive language more often than violent language, and the refusal to offer public assistance to these collectives.

Weak observer–level correlation and strong stimulus-level correlation between the McGurk effect and audiovisual speech-in-noise: A causal inference explanation

The McGurk effect is a widely used measure of multisensory integration during speech perception. Two observations have raised questions about the validity of the effect as a tool for understanding speech perception. First, there is high variability in perception of the McGurk effect across different stimuli and observers. Second, across observers there is low correlation between McGurk susceptibility and recognition of visual speech paired with auditory speech-in-noise, another common measure of multisensory integration. Using the framework of the causal inference of multisensory speech (CIMS) model, we explored the relationship between the McGurk effect, syllable perception, and sentence perception in seven experiments with a total of 296 different participants. Perceptual reports revealed a relationship between the efficacy of different McGurk stimuli created from the same talker and perception of the auditory component of the McGurk stimuli presented in isolation, both with and without added noise. The CIMS model explained this strong stimulus-level correlation using the principles of noisy sensory encoding followed by optimal cue combination within a common representational space across speech types. Because the McGurk effect (but not speech-in-noise) requires the resolution of conflicting cues between modalities, there is an additional source of individual variability that can explain the weak observer–level correlation between McGurk and noisy speech. Power calculations show that detecting this weak correlation requires studies with many more participants than those conducted to-date. Perception of the McGurk effect and other types of speech can be explained by a common theoretical framework that includes causal inference, suggesting that the McGurk effect is a valid and useful experimental tool.

The Semantic And Functional Characteristics Of Syntactic Stylistic Devices In The Example Of Repetition

This article aims to indicate the importance of syntactic stylistic devices mainly repetition in textual formulation that is based on the characteristics of the oral form of speech and syntactic stylistic devices which is expressed the structural meanings of the syntactic stylistic devices. In the aspects of this problem, it is important to analyze the role of syntactic stylistic devices in the formation of functional styles of the language, to justify the importance of syntactic stylistic devices in the presenting the main logic and contextual meanings in the literary texts. As analyzing any kind of literary text syntactic stylistic devices, that is repetition, chiasmus, stylistic inversion, gradation, ellipsis, antithesis, parallel construction, detached construction and enumeration have the greatest impact. However, according to results of analyzing, repetition is one of the widely used within these syntactic stylistic devices by contrast other models of speech.

Non-intrusive speech quality prediction based on the blind estimation of clean speech and the i-vector framework

Output-based instrumental speech quality assessment relies only on the received (processed) signal to predict quality. Such methods are called non-intrusive and are crucial in speech applications where reference clean signals are not accessible. In this paper, we propose a new non-intrusive instrumental quality measure based on the similarity between two i-vectors. As the reference clean signal is not available, the reference i-vector representation cannot be extracted directly from it. Therefore, we propose the use of a clean speech Gaussian mixture model to estimate the clean speech spectra from its degraded speech spectrum counterpart. Next, the two respective i-vector representations are extracted and either the cosine or Eucledian similarity metrics are computed as a correlate of speech quality. Here, the clean speech model is trained using RASTA-filtered mel-frequency cepstral coefficients extracted from a pool of clean speech files, thus allowing us to attain a model of clean spectrum characteristics. The proposed method is evaluated on noisy, reverberant, and enhanced speech conditions. Experimental results show the proposed system providing higher correlations with perceptual speech quality than several benchmark non-intrusive measures, especially for noisy and enhanced speech.

L2 Perception and Production of Japanese Lexical Pitch

Adults are known to have difficulties acquiring suprasegmental speech that involves pitch (f0) in a second language (L2) (Graham & Post, 2018; Hirata, 2015; Wang, Spence, Jongman & Sereno, 1999; Wong & Perrachione, 2007). Previous research has suggested that the perceived similarity between L1 and L2 phonology may influence how easily segmental speech is acquired, but this notion of ‘similarity’ may also apply to suprasegmental speech (So & Best, 2010; Wu, Munro & Wang, 2014). In this paper, the L2 acquisition of Japanese lexical pitch was assessed under a ‘Suprasegmental Similarity Account’, which is a theoretical framework inspired by previous models of segmental and suprasegmental speech (Best & Tyler, 2007; Flege, 1995; Mennen, 2015) to account for the L2 acquisition of word prosody. Eight adult native speakers of Japanese and eight adult English-native advanced learners of Japanese participated in a perception and production study of Japanese lexical pitch patterns. Both groups performed similarly in perception, but non-native speakers performed significantly worse in production, particularly for ‘unaccented’ Low–High–High patterns. These findings are discussed in light of the ‘Suprasegmental Similarity Account’.

TRANSFORMATION OF SCIENTIFIC TEXTS IN TEACHING RUSSIAN AS A FOREIGN LANGUAGE AT AN AGRICULTURAL UNIVERSITY

The article describes the forms and features of working with scientific texts when teaching Russian as a foreign language. A major role in this training is transforming syntactic structures: definition, classification of objects and phenomena, description of the subject composition, characteristics of the subject properties, characteristics, and changes substances. This work helps to remember case endings, as well as the development of new speech models. It is emphasized that texts containing scientific information play a special role in teaching scientific style. In these texts, special attention is given to post-text tasks where the student needs to complete sentences related to the test. The specificity of these sentences is that they have the same meaning, but a different structure than those contained in the text. It can be the tasks for choosing the case, replacing the noun with a verb, replacing the subordinate determinative, building a new phrase, which encourages students to show a certain creative approach to the task. It is suggested that an interesting type of work on the transformation of the text in the course of the development of written speech can be a task that requires changing the actual scientific or scientific-educational text in such a way that it becomes similar to popular science. It is concluded that the transformation of a scientific text when working with foreign students is of great importance in the formation of competence-based speech skills, since by transforming scientific texts, students improve their lexical and grammatical potential, which contributes to the solution of communicative tasks in the professional sphere.

TOLERANCE OF INCOHERENCE IN LAW, GRADED SPEECH ACTS AND ILLOCUTIONARY PLURALISM

AbstractOne of the most difficult challenges of mature legal systems is the need to balance the conflicting demands of stability and flexibility. The demand for flexibility is at odds with the principle of impartiality, which is considered a cornerstone of the rule of law. In the present article, I explore the way in which the law copes with this dilemma by developing the idea of tolerance of incoherence. I argue that tolerance of incoherence emerges from the interplay between the inferential and lexical-semantic rules that determine the meaning of legal speech acts. I base this argument on an inferential model of speech acts, which I develop through a discussion of graded speech acts, and on the idea that the use of speech acts is governed by multiple and potentially conflicting conventions. I show how this tolerance allows the law to resolve the tension between dynamism and traditionality, and discuss its sociological and moral implications.

New ways in hearing rehabilitation after cochlear implantation

EinleitungNach einer Cochleaimplantation (CI) ist i.d.R. ein Hör- und Sprachtraining erforderlich, das bislang entweder ambulant oder stationär im direkten Kontakt mit einem Therapeuten erfolgt. Bedingt durch technische Weiterentwicklungen, aber vor allem durch die zunehmende Nutzung digitaler Medien und dem Wunsch nach Mobilität und Selbständigkeit steigt der Bedarf an digitalen Trainingsprogrammen auch im Rahmen der Hörrehabilitation.Material und MethodenIn einem ersten Schritt wurden entsprechend den gängigen Hör- und Sprachverarbeitungsmodellen die wichtigsten Elemente eines Hörtrainings definiert und die Kernbausteine für eine computerbasierte teletherapeutische Hörplattform bestimmt. Dabei wurden lerntherapeutisch orientierte Elemente der Motivationsförderung sowie Mechanismen der Adaptivität, wie sie im Rahmen der logopädischen Therapie eingesetzt werden, berücksichtigt. In einem zweiten Schritt erfolgte die Entwicklung eines ersten Prototyps der teletherapeutischen Hörplattform „train2hear“ im Rahmen eines interdisziplinären Forschungsprojektes.ErgebnisseKernbaustein des vorgestellten „train2hear“ Hörtrainings ist eine an die „International Classification of Functioning, Disability, and Health“ (ICF) angelehnte Eingangsanalyse, aufgrund derer ein individueller Trainingsplan erstellt wird. Verschiedene Adaptivitätsmechanismen dienen dazu, den Schwierigkeitsgrad kontinuierlich an den individuellen Lernfortschritt des Nutzers anzupassen. Ein teletherapeutisches Videotool ermöglicht den Austausch mit dem Therapeuten vor Ort.Diskussion/FazitDie vorgestellte Plattform „train2hear“ stellt einen ersten deutschsprachigen Prototyp einer computerbasierten Hörrehabilitation dar. Dieser muss nun im Rahmen einer klinischen Studie im Hinblick auf die Anwendbarkeit und Effektivität evaluiert und weiterentwickelt werden.