Utterance Boundaries Research Articles

Markers of schizophrenia at the prosody/pragmatics interface. Evidence from corpora of spontaneous speech interactions.

The speech of individuals with schizophrenia exhibits atypical prosody and pragmatic dysfunctions, producing monotony. The paper presents the outcomes of corpus-based research on the prosodic features of the pathology as they manifest in real-life spontaneous interactions. The research relies on a corpus of schizophrenic speech recorded during psychiatric interviews (CIPPS) compared to a sampling of non-pathological speech derived from the LABLITA corpus of spoken Italian, which has been selected according to comparability requirements. Corpora has been intensively analyzed in the Language into Act Theory (L-AcT) frame, which links prosodic cues and pragmatic values. A cluster of linguistic parameters marked by prosody has been considered: utterance boundaries, information structure, speech disfluency, and prosodic prominence. The speech flow of patients turns out to be organized into small chunks of information that are shorter and scarcely structured, with an atypical proportion of post-nuclear information units (Appendix). It is pervasively scattered with silences, especially with long pauses between utterances and long silences at turn-taking. Fluency is hindered by retracing phenomena that characterize complex information structures. The acoustic parameters that give rise to prosodic prominence (f0 mean, f0 standard deviation, spectral emphasis, and intensity variation) have been measured considering the pragmatic roles of the prosodic units, distinguishing prominences within the illocutionary units (Comment) from those characterizing Topic units. Patients show a flattening of the Comment-prominence, reflecting impairments in performing the illocutionary activity. Reduced values of spectral emphasis and intensity variation also suggest a lack of engagement in communication. Conversely, Topic-prominence shows higher values for f0 standard deviation and spectral emphasis, suggesting effort when defining the domain of relevance of the illocutionary force. When comparing Topic and Comment-prominences of patients, the former consistently exhibit higher values across all parameters. In contrast, the non-pathological group displays the opposite pattern.

CLASSIC Utterance Boundary: A Chunking‐Based Model of Early Naturalistic Word Segmentation

AbstractWord segmentation is a crucial step in children's vocabulary learning. While computational models of word segmentation can capture infants’ performance in small‐scale artificial tasks, the examination of early word segmentation in naturalistic settings has been limited by the lack of measures that can relate models’ performance to developmental data. Here, we extended CLASSIC (Chunking Lexical and Sublexical Sequences in Children; Jones et al., 2021), a corpus‐trained chunking model that can simulate several memory and phonological and vocabulary learning phenomena to allow it to perform word segmentation using utterance boundary information, and we have named this extended version CLASSIC utterance boundary (CLASSIC‐UB). Further, we compared our model to the performance of children on a wide range of new measures, capitalizing on the link between word segmentation and vocabulary learning abilities. We showed that the combination of chunking and utterance‐boundary information used by CLASSIC utterance boundary allowed a better prediction of English‐learning children's output vocabulary than did other models.

The Sociolinguistic Variation of (āʔ) in Hatim

This sociolinguistic study investigates the variation in the pronunciation of the variable (āʔ) in the variety of Hatim, north-west Jordan. This variable is studied in relation to five factors: age, gender, education, outside contact and following linguistic environment. Data were collected via informal sociolinguistic interviews. Sixty-four participants (30 males, 34 females) were recorded spontaneously by the researchers (duration=2227 minutes). Two variants for the variable (āʔ) were found in the variety of Hatim: the tradition variant [ā] and the innovative one [āʔ]. The Rbrul analysis of the data shows that only outside contact, gender and following linguistic environment are statistically significant. Put differently, participants with high contact with other speech communities used the innovative variant more than those with low contact. As it is the case in most speech communities, female participants favoured the innovative variant over the traditional one. In relation to the following linguistic environment, it was found that the innovative variant is favoured when it does not occur at utterance boundaries. There appear to be two competing forces in the variety of Hatim, one for the traditional Horani feature (i.e., dropping the final glottal stop) and another for an innovative feature (i.e., keeping the final glottal stop) that happened to coincide with the pronunciation in Modern Standard Arabic. It can safely be argued that the driving force behind this change in progress is not a reversion to Modern Standard Arabic. Rather, it is the present-day social prestige associated with the pronunciation of the final glottal stop that is pushing for this change.
 
 Received: 8 August 2022 / Accepted: 20 October 2022 / Published: 5 November 2022

The Analysis of Two Components English Statements with the Translated Morpheme КАРТ-

Abstract. Nowadays the technical students’ interest in linguistic studies during English language classes in general and concerning land management in particular is being actualized in the context of defining metacommunicative markers as utterance boundaries for the formation of a terminological compound nest.The purpose of this article is to describe the practical experience of analyzing the features of two-component phrases with translated morpheme “карт-“ (hereinafter – DSMK) being based on the analysis of foreign scientific sources on land management according to the defined algorithm. The results of the study are presented in seven conclusions. The first conclusion is the expediency of students' analysis of phrases on land management according to the defined algorithm: to review the sources for the presence of the most common morphemes with high derivation potential and to be convinced of the expediency of studying a certain selected morpheme, to write down existing Ukrainian phrases with the required morpheme, assigning a specific number, spread terminological foreign phrases by blocks by structural features, to analyze according to other related characteristics for further generalization-collection of information from each block (according to the reliability of translation, the presence of proper names and numerical values, in relation to the land management industry, by the correspondence of the number of components of the Ukrainian and English variants and by the frequency of use in the educational texts and own statements of future land managers). The second conclusion is the expediency of taking into consideration the phrases with the translated morpheme “карт-“, as 108 varieties of stable English clichés were found. The third conclusion is the assignment of two-component attributive-nominative English phrases without additional conjunctions or prepositions to those that have an extremely high derivational potential to attract the translated card morpheme - in 99 varieties with a parallel assignment of two-component clichés with an additional preposition of or in / on connectors to those that have low derivation potential in only 9 varieties. The fourth conclusion is the assertion that the majority of DSMK with a significant quantitative advantage deals with the narrow specialization in cartography in contrast to the indirect being used in general thematic texts. The fifth conclusion is that the DSMK is more prone to the expected translation (74:34 ratio), although more than a third of these clichés should be guided by official translation options. The sixth conclusion is to determine the very low relevance of DSMK to proper names and numerical values and to a slightly higher percentage of DSMK with different number of components in the Ukrainian and English versions of the translation. The seventh conclusion is the fixation of the presence of DSMK in the educational texts and own statements of future land managers in the ratio 27:81, which proves the sufficient frequency of use of the translated morpheme “ карт-“ in regular phrases in the professional speech of future land managers, but not decisive. The prospect of further research is to describe the experience of the analysis of multicomponent constant clichés with a translated morpheme “карт-“ according to the proposed abovenamed algorithm.

Word-based largest chunks for Agreement Groups processing: Cross-linguistic observations

The present study reports results from a series of computer experiments seeking to combine word-based Largest Chunk (LCh) segmentation and Agreement Groups (AG) sequence processing. The AG model is based on groups of similar utterances that enable combinatorial mapping of novel utterances. LCh segmentation is concerned with cognitive text segmentation, i.e. with detecting word boundaries in a sequence of linguistic symbols. Our observations are based on the text of Le petit prince (The little prince) by Antoine de Saint-Exupéry in three languages: French, English, and Hungarian. The data suggest that word-based LCh segmentation is not very efficient with respect to utterance boundaries, however, it can provide useful word combinations for AG processing. Typological differences between the languages are also reflected in the results.

A Study on modality of the utterance boundaries―Focusing on epistemic modality Endings―

A Study on modality of the utterance boundaries―Focusing on epistemic modality Endings―

How Transitional Probabilities and the Edge Effect Contribute to Listeners’ Phonological Bootstrapping Success

ABSTRACTMuch of what we know about the development of listeners’ word segmentation strategies originates from the artificial language-learning literature. However, many artificial speech streams designed to study word segmentation lack a salient cue found in all natural languages: utterance boundaries. In this study, participants listened to a speech-stream containing one of three sets of word boundary cues: transitional probabilities between syllables (TP Condition), silences marking utterance boundaries (UB Condition), or a combination of both cues (TP + UB Condition). Recognition of the trained words and rule words (words not in language, but conforming to its phonotactic structure) was tested. Participants performed equally well in the TP + UB and UB Conditions, scoring above chance on both trained and rule words. Performance in the TP condition, however, was at chance. Our results suggest that attention to UBs is a particularly effective strategy for finding words in speech, possibly providing a language-general solution to the word segmentation problem.

Predicting pauses in L1 and L2 speech: the effects of utterance boundaries and word frequency

AbstractThis paper compares the distribution of silent and filled pauses in first (L1) and second language (L2) speech. The occurrence of pauses of 52 L2 and 18 L1 Dutch speakers was evaluated with respect to utterance boundaries and word frequency. We found that L2 speakers paused more often than L1 speakers within utterances; but not between utterances. Similarly, only within utterances, L2 pauses were longer than L1 pauses. Regarding word frequency, both L1 and L2 speakers are more likely to pause before lower frequency words as compared to higher frequency words. These findings imply that L1 and L2 speakers’ production processes may be similar in that (1) pauses at utterance boundaries are used for conceptual planning mostly and (2) lexical retrieval difficulties are comparable for L1 and L2 speakers. These findings furthermore imply that when using fluency for L2 testing, pause locations must be taken into account.

전사텍스트를 이용한 반자동 레이블링 구현

In transcription for spoken language research, labeling is a work linking text-represented utterance to re- corded speech. Most existing labeling tools have been working manually. Semi-automatic labeling we are proposing consists of automation module and manual adjustment module. Automation module extracts voice boundaries utilizing G.Saha's algorithm, and predicts utterance boundaries using the number and length of utterance which established utterance text. For maintaining existing manual tool's accuracy, we provide man- ual adjustment user interface revising the auto-labeling utterance boundaries. The implemented tool of our semi-automatic algorithm speed up to 27% than existing manual labeling tools.

Relações entre pausas e constituintes prosódicos na fala de crianças com desenvolvimento típico de linguagem

Objetivo (1) Verificar a distribuição de pausas não hesitativas e hesitativas em entrevistas com crianças em desenvolvimento típico de linguagem; (2) verificar, nas pausas não hesitativas, sua relação com limites de enunciados fonológicos e de frases entonacionais; (3) verificar, nas pausas hesitativas, sua distribuição entre início e interior de enunciados. Métodos As pausas foram extraídas de amostras de fala de quatro crianças com desenvolvimento típico de linguagem e idades entre quatro e cinco anos que, em 2011, frequentavam o nível Infantil II de uma Escola Municipal de Educação Infantil de Marília. As amostras de fala versavam sobre temas trabalhados em dez propostas pedagógicas desenvolvidas na sala de aula das crianças. Posteriormente, foi realizada, pela pesquisadora, uma inspeção de outiva (complementada pelo julgamento perceptual-auditivo de um grupo de cinco juízes) de cada um dos arquivos, a fim de se identificar os pontos de pausas. Resultados Não foi verificada diferença estatística na distribuição entre pausas hesitativas e não hesitativas; foi verificada tendência marginal de maior ocorrência de pausas em limites de frases entonacionais, do que em limites de enunciados fonológicos; não houve diferença estatística na distribuição entre pausas hesitativas em início e em interior de enunciados. Conclusão A instabilidade na distribuição das pausas é um importante recurso linguístico de observação dos domínios prosódicos que se mostram como mais, ou menos, dominados pelas crianças na aquisição da linguagem.

Prosody in a communication system developed without a language model.

Prosody, he "music" of language, is an important aspect of all natural languages, spoken and signed. We ask here whether prosody is also robust across learning conditions. If a child were not exposed to a conventional language and had to construct his own communication system, would that system contain prosodic structure? We address this question by observing a deaf child who received no sign language input and whose hearing loss prevented him from acquiring spoken language. Despite his lack of a conventional language model, this child developed his own gestural system. In this system, features known to mark phrase and utterance boundaries in established sign languages were used to consistently mark the ends of utterances, but not to mark phrase or utterance internal boundaries. A single child can thus develop the seeds of a prosodic system, but full elaboration may require more time, more users, or even more generations to blossom.

Telephone Speech Endpoint Detection using Mean-Delta Feature

Abstract In the study the efficiency of three features for trajectory-based endpoint detection is experimentally evaluated in the fixed-text Dynamic Time Warping (DTW) - a based speaker verification task with short phrases of telephone speech. The employed features are Modified Teager Energy (MTE), Energy-Entropy (EE) feature and Mean-Delta (MD) feature. The utterance boundaries in the endpoint detector are provided by means of state automaton and a set of thresholds based only on trajectory characteristics. The training and testing have been done with noisy telephone speech (short phrases in Bulgarian language with length of about 2 s) selected from BG-SRDat corpus. The results of the experiments have shown that the MD feature demonstrates the best performance in the endpoint detection tests in terms of the verification rate.

The edge factor in early word segmentation: utterance-level prosody enables word form extraction by 6-month-olds.

Past research has shown that English learners begin segmenting words from speech by 7.5 months of age. However, more recent research has begun to show that, in some situations, infants may exhibit rudimentary segmentation capabilities at an earlier age. Here, we report on four perceptual experiments and a corpus analysis further investigating the initial emergence of segmentation capabilities. In Experiments 1 and 2, 6-month-olds were familiarized with passages containing target words located either utterance medially or at utterance edges. Only those infants familiarized with passages containing target words aligned with utterance edges exhibited evidence of segmentation. In Experiments 3 and 4, 6-month-olds recognized familiarized words when they were presented in a new acoustically distinct voice (male rather than female), but not when they were presented in a phonologically altered manner (missing the initial segment). Finally, we report corpus analyses examining how often different word types occur at utterance boundaries in different registers. Our findings suggest that edge-aligned words likely play a key role in infants’ early segmentation attempts, and also converge with recent reports suggesting that 6-month-olds’ have already started building a rudimentary lexicon.

Formant Transitions in Varied Utterance Positions

Aim: Acoustic characteristics associated with varied utterance positions were examined to understand the acoustic consequences of potential articulatory changes near utterance boundaries. Methods: Second formant transition characteristics, including transition duration (ms), transition extent (Hz), and derived slope of transition (Hz/ms), of 12 healthy speakers of American English were examined for two diphthong transitions in sew and sigh and one consonant-vowel transition in bee in utterance-initial, utterance-final, and utterance-end positions. Speakers performed a task of contrastive stress variation that served to demonstrate the changeability of acoustic characteristics as an index of articulatory change in shaping the vocal tract. Results: Contrastive stress, as compared to words spoken without increased stress, was associated with longer transition duration, greater transition extent, and a decreased slope. Although some utterance position effects were present, no systematic differences consistent with boundary strengthening or declination were found. Conclusion: Findings suggest that varied utterance positions may be associated with stimulus-dependent variation in articulatory changes that is reflected in the acoustic output. These results indicate the need to further understand the construct of utterance-level speech materials, such as carrier phrases, in clinical practice and research.

Tracking Talking: Dual Task Costs of Planning and Producing Speech for Young versus Older Adults

ABSTRACT A digital pursuit rotor was used to monitor speech planning and production costs by time-locking tracking performance to the auditory wave form produced as young and older adults were describing someone they admire. The speech sample and time-locked tracking record were segmented at utterance boundaries and multilevel modeling was used to determine how utterance-level predictors such as utterance duration or sentence grammatical complexity and person-level predictors such as speaker age or working memory capacity predicted tracking performance. Three models evaluated the costs of speech planning, the costs of speech production, and the costs of speech output monitoring. The results suggest that planning and producing propositionally dense utterances is more costly for older adults and that older adults experience increased costs as a result of having produced a long, informative, or rapid utterance.

Enhanced voice activity detection using acoustic event detection and classification

We examine user-friendly voice interface that requires the hands-free speech acquisition in the continuously listening environment. The traditional voice activity detection (VAD) algorithms cannot successfully identify potential acoustic event sounds from speech. This makes the speech recognition system frequently or incorrectly activated. In this paper, we propose a novel voice activity detection technique that consists of two major modules: 1) classification and 2) detection module. In the classification module, we label the successive audio segments based on the training models. Then, in the detection module, we remove the acoustic event sounds and make decision of the explicit utterance boundary from the input audio stream. As a result, the proposed technique enables the efficient operation of speech recognition in the continuously listening environment without any touch and/or key input. Experiments in a real-world environment and performance comparison with state-of-the-art techniques are conducted to demonstrate the effectiveness of the proposed technique.

Listener reliability in assigning utterance boundaries in children's spontaneous speech

ABSTRACTResearch and clinical practices often rely on an utterance unit for spoken language analysis. This paper calls attention to the problems encountered when identifying utterance boundaries in young children's spontaneous conversational speech. The results of a reliability study of utterance boundary assignment are described for 20 females with graduate professional education in speech–language pathology. They judged audiorecorded spontaneous speech samples that were elicited from 4-year-old children. Their agreement with each other (interobserver) was significantly lower than their self-agreement (intraobserver). Interobserver agreement varied with the length/grammatical complexity of response turns, and the number and types of selective boundary cues presented in the speech stimuli. The findings have research and clinical implications for the utterance as a conceptual notion and its use constraints.

Words in puddles of sound: modelling psycholinguistic effects in speech segmentation

There are numerous models of how speech segmentation may proceed in infants acquiring their first language. We present a framework for considering the relative merits and limitations of these various approaches. We then present a model of speech segmentation that aims to reveal important sources of information for speech segmentation, and to capture psycholinguistic constraints on children's language perception. The model constructs a lexicon based on information about utterance boundaries and deduces phonotactic constraints from the discovered lexicon. Compared to other models of speech segmentation, our model performs well in terms of accuracy, computational tractability and the number of components of the model. Finally, our model also reflects the psycholinguistic effects of language learning, in terms of the early advantage for segmentation provided by the child's name, and by revealing the overlap in usefulness of information for segmentation and for grammatical categorization of the language.

On the acoustic correlates of high and low nuclear pitch accents in American English

Earlier findings in Shue et al. (2007, 2008) raised questions about the alignment of nuclear pitch accents in American English, which are addressed here by eliciting both high and low pitch accents in two different target words in several different positions in a single-phrase utterance (early, late but not final, and final) from 20 speakers (10 male, 10 female). Results show that the F 0 peak associated with a high nuclear pitch accent is systematically displaced to an earlier point in the target word if that word is final in the phrase and thus bears the boundary-related tones as well. This effect of tonal crowding holds across speakers, genders and target words, but was not observed for low accents, adding to the growing evidence that low targets behave differently from highs. Analysis of energy shows that, across target words and genders, the average energy level of a target word is greatest at the start of an utterance and decreases with increasing proximity to the utterance boundary. Duration measures confirms the findings of existing literature on main-stress-syllable lengthening, final syllable lengthening, and lengthening associated with pitch accents, and reveals that final syllable lengthening is further enhanced if the final word also carries a pitch accent. Individual speaker analyses found that while most speakers conformed to the general trends for pitch movements there were 2/10 male and 1/10 female speakers who did not. These results show the importance of taking into account prosodic contexts and speaker variability when interpreting correlates to prosodic events such as pitch accents.

The use of prosodic markers to indicate utterance boundaries in American Sign Language interpretation

The use of prosodic markers to indicate utterance boundaries in American Sign Language interpretation