Speaker Change Research Articles

Speaker variability in spoken word recognition: Evidence from non-native listeners

Non-native listeners face many challenges during spoken word recognition. On the lexical level, the vocabulary size of non-native listeners is usually reduced relative to native listeners. Lexical competition can arise from the lexicon in both the native and the non-native language during speech perception. How do non-native listeners resolve speaker variability, a challenge that is presumably non-lexical in spoken word recognition? This study aims to explore non-native spoken word recognition in the face of speaker variability. Mandarin speakers using English as a second or foreign language are recruited to listen to pairs of English words (e.g., cat—dog). Lexical decision time and accuracy on the second item in a pair (e.g., dog) are measured. The pairs are either spoken by the same or different native speakers of American English to examine the effect of speaker change on reaction time and accuracy in a short-term priming paradigm. Non-native listeners’ pattern of response is explored as a function of the amount of priming received against speaker variability.

Applied Nanotechnology and Nanoscience International Conference 2017 (ANNIC 2017)

PrefaceIntroduction:The Applied Nanotechnology and Nanoscience International Conference is an international meeting organized in a different European country each year. The Chairs Committee and Plenary Speakers change each year in order to enrich the content and provide the participants with different views.After the success of Paris (2015) and Barcelona (2016), the Applied Nanotechnology and Nanoscience International Conference 2017 was held in Rome, on October 18th -20th.

Pathways to adversity and speaker affectedness: On the emergence of unaccusative ‘give’ constructions in Chinese

Abstract In this paper, we examine the characteristics of unaccusative ‘give’ constructions in Chinese, and additionally identify the pathways for their emergence in some Chinese dialects, in particular Southern Min and Mandarin varieties. In this paper, the terms dialect and variety are sometimes used interchangeably, with the term variety being the more general term that can also include variations within dialects. We distinguish between Type 1 and Type 2 unaccusative ‘give’ constructions, the former involving reversible ‘escape’-type intransitive predicates, and the latter irreversible ‘die’-type intransitive predicates. Type 1 constructions are attested in many Chinese varieties, such as Mandarin, Min, Wu, Hui, Hakka and Cantonese, whereas Type 2 constructions are more rare and thus far are mainly attested in Southern Min and some Mandarin varieties. Two major pathways in the development of unaccusative ‘give’ constructions are identified in this paper, namely, the causative pathway and the passive-mediated pathway. Our analysis also traces how the unaccusative ‘give’ construction develops into a marker of adversity and speaker affectedness. The findings of this study have implications for understanding the relationship between changes in valence (i.e., the number of core arguments that are profiled in a given construction) and speaker’s subjective stance.

Predicting Intelligibility Gains in Individuals With Dysarthria From Baseline Speech Features.

Across the treatment literature, behavioral speech modifications have produced variable intelligibility changes in speakers with dysarthria. This study is the first of two articles exploring whether measurements of baseline speech features can predict speakers' responses to these modifications. Fifty speakers (7 older individuals and 43 speakers with dysarthria) read a standard passage in habitual, loud, and slow speaking modes. Eighteen listeners rated how easy the speech samples were to understand. Baseline acoustic measurements of articulation, prosody, and voice quality were collected with perceptual measures of severity. Cues to speak louder and reduce rate did not confer intelligibility benefits to every speaker. The degree to which cues to speak louder improved intelligibility could be predicted by speakers' baseline articulation rates and overall dysarthria severity. Improvements in the slow condition could be predicted by speakers' baseline severity and temporal variability. Speakers with a breathier voice quality tended to perform better in the loud condition than in the slow condition. Assessments of baseline speech features can be used to predict appropriate treatment strategies for speakers with dysarthria. Further development of these assessments could provide the basis for more individualized treatment programs.

Turn-Taking in Colloquial Indonesian

Participants in a conversation follow certain basic rules in obtaining turns. Some questions addressed in this paper are, firstly, how turns are allocated in Indonesian conversation. Secondly, what are the signals given by a current speaker willing to yield to a next speaker (i.e. turn-yielding signals). Thirdly, what repair mechanism is employed by the participants when the turn-taking procedures are broken. The data used in this study were taken from an informal conversation of four Indonesians from Jakarta. The data were collected by using a non-participatory observation with a recording technique. The analysis shows that the turn-taking system in the Indonesian language confirms the model of turn-taking system developed by Sacks et al (1974), covering the occurrence and the recurrence of speaker change, the overwhelming tendency for one party to talk at a time,the common occurrences of more than one speaker at a time, and the common switches of speakers with no gapsand no overlaps.

Adaptive speaker diarization of broadcast news based on factor analysis

The introduction of factor analysis techniques in a speaker diarization system enhances its performance by facilitating the use of speaker specific information, by improving the suppression of nuisance factors such as phonetic content, and by facilitating various forms of adaptation. This paper describes a state-of-the-art iVector-based diarization system which employs factor analysis and adaptation on all levels. The diarization modules relevant for this work are: the speaker segmentation which searches for speaker boundaries and the speaker clustering which aims at grouping speech segments of the same speaker. The speaker segmentation relies on speaker factors which are extracted on a frame-by-frame basis using eigenvoices. We incorporate soft voice activity detection in this extraction process as the speaker change detection should be based on speaker information only and we want it to disregard the non-speech frames by applying speech posteriors. Potential speaker boundaries are inserted at positions where rapid changes in speaker factors are witnessed. By employing Mahalanobis distances, the effect of the phonetic content can be further reduced, which results in more accurate speaker boundaries. This iVector-based segmentation significantly outperforms more common segmentation methods based on the Bayesian Information Criterion (BIC) or speech activity marks. The speaker clustering employs two-step Agglomerative Hierarchical Clustering (AHC): after initial BIC clustering, the second cluster stage is realized by either an iVector Probabilistic Linear Discriminant Analysis (PLDA) system or Cosine Distance Scoring (CDS) of extracted speaker factors. The segmentation system is made adaptive on a file-by-file basis by iterating the diarization process using eigenvoice matrices adapted (unsupervised) on the output of the previous iteration. Assuming that for most use cases material similar to the recording in question is readily available, unsupervised domain adaptation of the speaker clustering is possible as well. We obtain this by expanding the eigenvoice matrix used during speaker factor extraction for the CDS clustering stage with a small set of new eigenvoices that, in combination with the initial generic eigenvoices, models the recurring speakers and acoustic conditions more accurately. Experiments on the COST278 multilingual broadcast news database show the generation of significantly more accurate speaker boundaries by using adaptive speaker segmentation which also results in more accurate clustering. The obtained speaker error rate (SER) can be further reduced by another 13% relative to 7.4% via domain adaptation of the CDS clustering.

The development of gaze to a speaking face.

When a speaker talks, the visible consequences of what they are saying can be seen. Listeners are influenced by this visible speech both in a noisy listening environment and even when auditory speech can easily be heard. While visible influence on heard speech has been reported to increase from early to late childhood, little is known about the mechanism that underlies this developmental trend. One possible account of developmental differences is that looking behavior to the face of a speaker changes with age. To examine this possibility, the gaze to a speaking face was examined in children from 5 to 10 yrs of age and adults. Participants viewed a speaker's face in a range of conditions that elicit looking: in a visual only (speech reading) condition, in the presence of auditory noise (speech in noise) condition, and in an audiovisual mismatch (McGurk) condition. Results indicate an increase in gaze on the face, and specifically, to the mouth of a speaker between the ages of 5 and 10 for all conditions. This change in looking behavior may help account for previous findings in the literature showing that visual influence on heard speech increases with development.

GMM-UBM 기반 KL 거리를 활용한 화자변화 검증에 대한 연구

본 논문에서는 기존의 BIC(Bayesian Information Criterion) 기반 화자변화의 성능 향상을 위하여 GMM-UBM(Gaussian Mixture Model-Universal Background Model) 기반 KL(Kullback Leibler) 거리를 활용한 화 자변화 검증을 제안하였다. 정보량의 차이에 민감한 기존의 BIC 기반 화자변화검출 알고리즘을 상대적으로 정보량 차이에 견인한 KL 거리 알고리즘으로 검증하였고, 정보량의 비대칭을 보상하기 위해서 GMM-UBM을 활용하였다. 기존의 BIC 기반 화자변화 검출은 1단계로 비유사도 d가 양수인 구간의 국소 최댓값인 지점을 화자변화 후보지점으 로 검출하였고, 2단계로 검출된 화자변화 후보지점 중 ΔBIC가 양수인 지점을 화자변화지점으로 결정하였다. 본 논 문에서는 BIC 기반 화자변화 검출에 의해 결정된 화자변화지점에 대하여 GMM-UBM 기반 KL 거리 D가 문턱치 (threshold)보다 높은 지점을 최종 화자변화 지점으로 검증하였다. 실험결과, MDR(Missed Detection Rate)이 0인 조건에서 문턱치 0.028일 때 FAR(False Alarm Rate) 60.4%로 성능이 향상되었다.

VIENAS IŠ TRIJŲ: SEIMO PIRMININKO STATUSAS IR VAIDMUO

Šiame straipsnyje, remiantis neoinstitucionalizmo ir politinės lyderystėsteorijų nuostatomis, analizuojami parlamento pirmininko pareigybėsstatusas, vaidmuo ir reikšmė parlamento institucionalizacijos kontekste, pristatomi mokslinėje literatūroje nurodomi šios pareigybės tipai, pagrindžiamosteorinės prielaidos analizuoti Seimo Pirmininko pareigybę. Remiantisišskirtais kriterijais, analizuojami Lietuvos Respublikos Seimo Pirmininkovaidmuo ir reikšmė politikos procese, per Lietuvos nepriklausomybės istorijąsusiformavęs šios pareigybės tipas, jis lyginamas su atitinkamais Vakarųvalstybių modeliais.

Switching of auditory attention in “cocktail-party” listening: ERP evidence of cueing effects in younger and older adults

Verbal communication in a “cocktail-party situation” is a major challenge for the auditory system. In particular, changes in target speaker usually result in declined speech perception. Here, we investigated whether speech cues indicating a subsequent change in target speaker reduce the costs of switching in younger and older adults. We employed event-related potential (ERP) measures and a speech perception task, in which sequences of short words were simultaneously presented by four speakers. Changes in target speaker were either unpredictable or semantically cued by a word within the target stream. Cued changes resulted in a less decreased performance than uncued changes in both age groups. The ERP analysis revealed shorter latencies in the change-related N400 and late positive complex (LPC) after cued changes, suggesting an acceleration in context updating and attention switching. Thus, both younger and older listeners used semantic cues to prepare changes in speaker setting.

Speaker Diarization and Linking of Meeting Data

Finding who spoke when in a collection of recordings, with speakers being uniquely identified across the database, is a challenging task. In this scenario, reasonable computing times and acoustic variation across recordings remain two major concerns to address in state-of-the-art speaker diarization systems. This paper extends prior work on diarizing large speech datasets using algorithms that scale well with increasing amounts of data while compensating for across-recording variability. We follow a two-stage approach performing speaker diarization and speaker linking, the former focusing on local within-recording speaker changes and the latter focusing on global speaker changes across the database. In this study, we explore how these two modules interact with each other, while proposing a diarization fusion approach that prevents diarization errors from propagating to the linking stage. We further explore the diarization fusion for speaker linking using different linking strategies and speaker modeling variants. Evaluation is performed on single distant microphone data from the augmented multiparty interaction corpus show the effectiveness of the fusion approach after speaker linking and intersession variability modeling via joint factor analysis.

Emotional speech feature normalization and recognition based on speaker-sensitive feature clustering

In this paper we propose a feature normalization method for speaker-independent speech emotion recognition. The performance of a speech emotion classifier largely depends on the training data, and a large number of unknown speakers may cause a great challenge. To address this problem, first, we extract and analyse 481 basic acoustic features. Second, we use principal component analysis and linear discriminant analysis jointly to construct the speaker-sensitive feature space. Third, we classify the emotional utterances into pseudo-speaker groups in the speaker-sensitive feature space by using fuzzy k-means clustering. Finally, we normalize the original basic acoustic features of each utterance based on its group information. To verify our normalization algorithm, we adopt a Gaussian mixture model based classifier for recognition test. The experimental results show that our normalization algorithm is effective on our locally collected database, as well as on the eNTERFACE'05 Audio-Visual Emotion Database. The emotional features achieved using our method are robust to the speaker change, and an improved recognition rate is observed.

A RÁDIÓBAN ELHANGZOTT DIALÓGUSOK JELLEMZŐI

A tanulmány társalgáselemzési szempontokat figyelembe véve elemzi az Újvidéki Rádió magyar nyelvű műsoraiban elhangzott beszélgetések jellemzőit, de kitér pragmatikai és nyelvi sajátságokra is. Dialógusrészletek idézésével bemutatja, miként valósulnak meg a társalgás forgatókönyvének elemei, többek között a beszédbe elegyedés, a beszélőváltás és a beszélgetés befejezése. Az elemzés alapján megállapítható, hogy nemcsak a kérdések, hanem a megjegyzések és tényközlések is szervezik a dialógusokat. A mindennapi dialógusoktól eltérően a formális helyzetben folytatott dialógusokban csak ritkán van témaváltás. Bemutatja, hogyan történik a visszatérés az eredeti témához. A rádióban elhangzott spontán dialógusokban is vannak nyelvi hibák. A példák szemléltetik, hogyan történik ezek javítása. A társalgás kontextusa magában foglalja a szituációs kontextust, a cselekvés kontextusát és a tematikus kontextust. Ezek összetevői többek között a térre és az időre utalásra, valamint a résztvevők közötti viszony jelölése. A tanulmány bemutatja ezeknek az összetevőknek a megjelenését és nyelvi kifejezőeszközeit.

Appeals to Ethos and Pathos in Post-Crisis Communication: A Case Study of Press Conferences Handling Tianjin Blasts

<p>This study focuses on the rhetorical appeals in post-crisis communication, and analyzes the influence of dynamically changing context on the speakers’ choice of rhetorical means. Aristotle’s three important rhetorical appeals-ethos, pathos and logos are investigated in the transcript of 14 press conferences handling Tianjin blasts. The changes in speakers’ rhetorical appeals are explored in relation to the changing context in the evolving crisis. In post-crisis press conferences, the speakers most frequently used appeals to ethos and pathos to persuade. Specifically, when appealing to ethos, the speakers usually manipulated discursive resources to construct their credibility, expertise, and similarity with the audience; when appealing to pathos, such emotions of the public as the feelings of depression, need for appreciation, and expectation for a thorough investigation of the blasts were addressed. In addition, as the crisis unfolded, the context for communication was dynamically changing. There was a tendency for speakers to adapt their rhetorical appeals to the dynamically changing context.</p>

Listening through voices: Infant statistical word segmentation across multiple speakers.

To learn from their environments, infants must detect structure behind pervasive variation. This presents substantial and largely untested learning challenges in early language acquisition. The current experiments address whether infants can use statistical learning mechanisms to segment words when the speech signal contains acoustic variation produced by changes in speakers' voices. In Experiment 1, 8- and 10-month-old infants listened to a continuous stream of novel words produced by 8 different female voices. The voices alternated frequently, potentially interrupting infants' detection of transitional probability patterns that mark word boundaries. Infants at both ages successfully segmented words in the speech stream. In Experiment 2, 8-month-olds demonstrated the ability to generalize their learning about the speech stream when presented with a new, acoustically distinct voice during testing. However, in Experiments 3 and 4, when the same speech stream was produced by only 2 female voices, infants failed to segment the words. The results of these experiments indicate that low acoustic variation may interfere with infants' efficiency in segmenting words from continuous speech, but that infants successfully use statistical cues to segment words in conditions of high acoustic variation. These findings contribute to our understanding of whether statistical learning mechanisms can scale up to meet the demands of natural learning environments.

Second Dialect Acquisition: A Sociophonetic Perspective

AbstractMany people change aspects of their accent after moving to a new region. What kinds of changes are made, and why does it matter? Studies of second dialect acquisition (SDA) indicate that geographically mobile speakers change specific dialect features in ways that reflect the complex interaction of linguistic, social, and developmental factors in language use. This article reviews these findings from a sociophonetic perspective, paying particular attention to their theoretical implications, the methodological issues associated with studying SDA, and avenues for future research.

Communicating printed headings to the ear

Two experiments compared three methods of translating printed headings into an auditory format. In both experiments, college students listened to a text with instructions to stop the recording whenever they heard a heading and type the hierarchical level and exact wording of the heading. Listeners were poor at identifying headings and their levels if the headings were not distinguished from the rest of the text. However, listeners were very good at identifying headings if any method of signalling was used to distinguish headings and communicate their hierarchical level. The methods included: (1) tones preceding headings, (2) changes of speaker to indicate headings or (3) verbal labels preceding headings. Although all three signalling methods improved identification of a heading’s hierarchical level, the labelling method was the most effective means of communicating hierarchical level. Thus, the study identifies a simple method of effectively communicating headings in spoken text.Practitioner Summary:The study attempted to identify effective ways of communicating heading information in spoken text. College students listened to texts in order to detect headings and record their wording and hierarchical level. Performance was excellent when headings were preceded by verbal phrases that signalled the upcoming headings and their levels.

Phonetic and lexical influences on changes across the lifespan

The study considers longitudinal studies over several decades within the same individual in order to determine whether phonetic sound change takes place initially in more frequent words. The focus of the analysis was on the vowels of the Christmas Broadcasts by Queen Elisabeth II over several decades and in Alistair Cooke's Letter from America broadcasts. For the first of these, a re-analysis of the phonetic lowering of the vowel in the lexical set TRAP and of tensing of final lax vowel in HAPPY showed no effect of lexical frequency on sound change. The focus of analysis in the second was during a period in which the speaker was shown to acquire General American characteristics after emigrating to the United States from Britain, but then in later life to revert over a 5–10 year period back toward characteristics of British English Received Pronunciation. This reversion is shown to take place at a faster rate in lexically frequent words. While the evidence overall for lexically gradual changes is equivocal, the changes in both speakers are most appropriately modeled as a leveling due to variation in dialect contact: increasingly with middle class speakers for the Queen; decreasingly with American English speakers for Cooke.

Effects of Adventitious Acute Vocal Trauma: Relative Fundamental Frequency and Listener Perception

High voice users (individuals who demonstrate excessive or loud vocal use) are at risk for developing voice disorders. The objective of this study was to examine, both acoustically and perceptually, vocal changes in healthy speakers after an acute period of high voice use. Members of a university women's volleyball team (n=12) were recorded a week before (pre) and week after (post) the 10-week spring season; n=6 control speakers were recorded over the same time period for comparison. Speakers read four sentences, which were analyzed for relative fundamental frequency (RFF). Eight naïve listeners participated in an auditory-perceptual visual sort and rate (VSR) task, in which they rated each voice sample's overall severity and strain. No significant differences were found as a function of time point in the VSR ratings for the volleyball group. Onset cycle 1 RFF values were significantly lower (P=0.04) in the postrecordings of the volleyball participants compared with prerecordings, but there was no significant difference (P=0.20) in offset cycle 10 RFF values. Receiver operating characteristic analyses indicated moderate sensitivity and specificity of onset cycle 1 RFF for discrimination between the volleyball and control participants. Changes were not apparent in the control group as a function of time for either, onset cycle 1 RFF, offset cycle 10 RFF, or either vocal attribute. Onset cycle 1 RFF may be an effective marker for detecting vocal changes over an acute high voice use period of time before perceptual changes are noted.

What is the source of L1 attrition? The effect of recent L1 re-exposure on Spanish speakers under L1 attrition

The recent hypothesis that L1 attrition affects the ability to process interface structures but not knowledge representations (Sorace, 2011) is tested by investigating the effects of recent L1 re-exposure on antecedent preferences for Spanish pronominal subjects, using offline judgements and online eye-tracking measures. Participants included a group of native Spanish speakers experiencing L1 attrition (‘attriters’), a second group of attriters exposed exclusively to Spanish before they were tested (‘re-exposed’), and a control group of Spanish monolinguals. The judgement data shows no significant differences between the groups. Moreover, the monolingual and re-exposed groups are not significantly different from each other in the eye-tracking data. The results of this novel manipulation indicate that attrition effects decrease due to L1 re-exposure, and that bilinguals are sensitive to input changes. Taken together, the findings suggest that attrition affects online sensitivity with interface structures rather than causing a permanent change in speakers’ L1 knowledge representations.