Prosodic Word Level Research Articles

Prosodic Word Recursion in a Polysynthetic Language (Blackfoot; Algonquian)

This paper focuses on prosodic adjunction at the Prosodic Word level in a polysynthetic language. I argue that recursion at a depth of more than two levels can only be generated by a theory which requires exact correspondence between certain syntactic phrases and Prosodic Words. Such a theory is similar to Phonological Phrase correspondence in Match Theory, suggesting there is an underlying shared property between correspondence at the Prosodic Word and Phonological Phrase levels. In addition, this theory must include a constraint which prohibits recursive prosodic constituents in order to generate the attested typology of clitics across languages. The empirical focus is the prosodic structure of the verbal complex in Blackfoot (Algonquian; ISO 639-3: bla). Using phonotactic evidence I argue that the vP phase corresponds to a Prosodic Word, and that each prefix to the stem is a Prosodic Word adjunct. I then compare several theories of the syntax-prosody interface, including versions of Alignment Theory, Wrap Theory, and Match Theory. A subset of schematic candidates with one or two prefixes to a stem are used to determine which theories generate the attested typology of clitics as well as a multiply recursive Prosodic Word structure.

Word stress at utterance-final position

Abstract This study investigates the realization of the two most common word-level stress patterns in Hebrew, final and penultimate, at utterance-final position. Twenty-six disyllabic words that form minimal pairs, which differ only in their stress pattern, were embedded in 52 sentences. The mean values of three acoustic parameters—duration, F0, and intensity—were measured for vowels of the target words. Findings show that duration is significantly longer at stressed vowels, similar to previous findings on words at utterance-mid position. Lower intensity is assigned to the utterance-final vowels regardless of the stress pattern, but the degree of lowering does depend on the stress pattern. Finally, lower F0 values are found in the utterance-final vowels, but the degree of lowering is similar to both stress patterns. We conclude that duration is the main cue at the prosodic word level, while F0 is used by Hebrew speakers to cue higher prosodic units.

Personalized Spontaneous Speech Synthesis Using a Small-Sized Unsegmented Semispontaneous Speech

A systematic approach is proposed to synthesizing personalized spontaneous speech using a small-sized unsegmented speech corpus of the target speaker. First, an automatic segmentation algorithm is employed to segment and label a collected semispontaneous speech corpus of the target speaker. Then, a pretrained average voice model is adapted to the voice model of the target speaker by using the segmented data. A postfilter based on modulation spectrum is adopted to further improve the speaker similarity of the synthesized speech as well as alleviate the over-smoothing problem of the synthesized speech. For generating spontaneous speech, a smoothing method applied at the prosodic word level is proposed to improve speech fluency. For objective evaluation on spontaneous speech segmentation, the segmentation accuracy of the proposed method is superior to that of Viterbi-based forced alignment. The results of subjective listening test also show that the proposed method can improve the spontaneity and speaker similarity of the synthesized speech compared to the maximum likelihood linear regression based speaker adaptation method.

Discourse prosody planning in native (L1) and nonnative (L2) (L1-Bengali) English: a comparative study

This paper conducts a comparative study between L1 and L2 (L1 Bengali) English discourse level speech planning to investigate differences between L1 and L2 English speaker groups in the organization of discourse-level speech planning. For this purpose, English speech of 10 L1 English and 40 L1 Bengali speakers of the same discourse are analyzed in terms of using prosodic and acoustic cues by applying hierarchical discourse prosody framework. From this analysis, between-group differences in discourse level speech planning are found through the speech rate, locations of discourse boundary breaks as well as size and scope of speech planning and chunking units. Result of analysis shows that the speech rate of L1 English speakers is higher than that of L2 English speakers, L2 English speakers contain more break boundary than that of the L1 English speakers at every discourse level in the organization, which exhibit the fact that L2 English speakers use more intermediate chunking units and larger scale planning units than that of L1 English speakers. Between-group differences are also found through the analysis of phrase component at prosodic phrase level and accent component at the prosodic word level. These findings can be attributed to L2 English speakers' improper phrasing, improper word level prominence and the ambiguous difference between content words and function words. The study concludes that the deficiencies in English strategy for L1 Bengali speakers' discourse-level speech planning compared to L1 English speakers are due to the influence of L1 (Bengali) prosody at the L2 discourse level.

Candidate Expansion and Prosody Adjustment for Natural Speech Synthesis Using a Small Corpus

This study proposes a hybrid approach to natural-sounding speech synthesis based on candidate expansion, unit selection, and prosody adjustment using a small corpus. The proposed method is more specific to tonal language, in particular Mandarin. In conventional speech synthesis studies, the quality of synthesized speech depends heavily on the size of the speech corpus. However, it is highly time-consuming and labor-intensive to prepare a large labeled corpus. In this work, candidate expansion is proposed to retrieve potential candidates that are unlikely to be retrieved using only linguistic features. The optimal unit sequence is then obtained from the expanded candidates by using the proposed unit selection mechanism at the phoneme and prosodic word levels. Finally, a prosodic word-level prosody adjustment is proposed to improve the continuity and smoothness of the prosody of the synthesized speech. To evaluate the proposed method, the Tsing-Hua corpus of speech synthesis was adopted. The results of an objective evaluation demonstrate the effectiveness of candidate expansion and the improvement of the continuity and smoothness of the prosody of the synthesized speech. The results of a subjective evaluation also show the proposed system could synthesize the speech with improved quality and naturalness, in particular for a small-sized or resource-limited corpus.

Imitation and Naming in Childhood Apraxia of Speech

Imitation and object naming are the first steps in language acquisition. While in typical development, imitation is easier than naming, there is a lack of agreement about the role of imitation in language impairment in general, and in Childhood Apraxia of Speech (CAS), in particular. Hence, the aim of this study is to compare imitation and naming in CAS and to analyze the nature of errors according to the complexity of the word and across the three levels of the prosodic hierarchy: the segmental level, the syllabic level and the prosodic word level. Sixteen children diagnosed with CAS (average age 3;11) participated in the study. The data, collected from each child in the course of eight weekly meetings, are drawn from naming and imitating single words. The results indicate that imitation seems to be an easier task than naming; both tasks are more difficult as the words get longer. Fewer errors are seen on the prosodic level, while most of the errors occur on the segmental and syllabic levels.

Domains of H tone spreading and the noun class prefix in Xitsonga

This article reports results from an investigation of domains of H tone spreading in Xitsonga, a southern Bantu language. High (H) tone spreads into toneless syllables but it spreads only to the first syllable if a nominal root has an H tone. Kisseberth (1994) argues that domain structures created by the Pre-High Projection rule account for the patterns regarding the non-spreading of H tone in Xitsonga. The findings concur with Kisseberth's analysis of domain sensitive H tone spreading when nouns have one or two toneless syllables before an H-toned syllable. In this paper, Kisseberth's observation is extended by analysing H tone spreading in nouns with more than two toneless syllables before an H-tone syllable. Results show that H tone spreads only to the noun class prefix, which suggests that the noun class prefix is outside of the prosodic domain formed in nouns. I further propose that the Obligatory Contour Principle functions at the prosodic word level and that the noun class prefix of a noun phrase in Xitsonga is encliticised to a preceding word.

Head and facial gestures synthesis using PAD model for an expressive talking avatar

This paper proposes to synthesize expressive head and facial gestures on talking avatar using the three dimensional pleasure-displeasure, arousal-nonarousal and dominance-submissiveness (PAD) descriptors of semantic expressivity. The PAD model is adopted to bridge the gap between text semantics and visual motion features with three dimensions of pleasure-displeasure, arousal-nonarousal, and dominance-submissiveness. Based on the correlation analysis between PAD annotations and motion patterns derived from the head and facial motion database, we propose to build an explicit mapping from PAD descriptors to facial animation parameters with linear regression and neural networks for head motion and facial expression respectively. A PAD-driven talking avatar in text-to-visual-speech system is implemented by generating expressive head motions at the prosodic word level based on the (P, A) descriptors of lexical appraisal, and facial expressions at the sentence level according to the PAD descriptors of emotional information. A series of PAD reverse evaluation and comparative perceptual experiments shows that the head and facial gestures synthesized based on PAD model can significantly enhance the visual expressivity of talking avatar.

Intra-word inconsistency in apraxic Hebrew-speaking children

Intra-word inconsistency in a child is perceived as an indicator of speech impairment. Because the speech of typically developing children is highly variable, the extent and nature of the inconsistency must be defined when used as a diagnostic marker of speech impairment (McLeod, S., & Hewett, S. R. (2008). Variability in the production of words containing consonant clusters by typical 2- and 3-year-old children. Folia Phoniatrica et Logopaedica, 60(4), 163–172). In this paper, we study inconsistency with reference to the prosodic hierarchy (McCarthy, J. J., & Prince, A. S. (1996). Prosodic morphology 1986. Amherst, MA: University of Massachusetts. Retrieved April 15, 2010, from http://ruccs.rutgers.edu/pub/papers/pm86all.pdf), suggesting a new way to describe this phenomenon in childhood apraxia of speech (CAS). The prosodic hierarchy has been used in recent years to demonstrate the phonological development of typical and atypical populations. Sixteen children diagnosed with CAS (average age 3;11) participated in the study. The data, collected from each child in the course of eight weekly meetings, are drawn from naming single words. The results indicate that inconsistency is dominant for two prosodic levels, the segmental and the syllabic, while the prosodic word level was largely preserved.

Sensitivity to Visual Prosodic Cues in Signers and Nonsigners

Three studies are presented in this paper that address how nonsigners perceive the visual prosodic cues in a sign language. In Study 1, adult American nonsigners and users of American Sign Language (ASL) were compared on their sensitivity to the visual cues in ASL Intonational Phrases. In Study 2, hearing, nonsigning American infants were tested using the same stimuli used in Study I to see whether maturity, exposure to gesture, or exposure to sign language is necessary to demonstrate this type of sensitivity. Study 3 addresses nonsigners' and signers' strategies for segmenting Prosodic Words in a sign language. Adult participants from six language groups (3 spoken languages and 3 sign languages) were tested.The results of these three studies indicate that nonsigners have a high degree of sensitivity to sign language prosodic cues at the Intonational Phrase level and the Prosodic Word level; these are attributed to modality or'channel' effects of the visual signal.There are also some differences between signers' and nonsigners' sensitivity; these differences are attributed to language experience or language-particular constraints.This work is useful in understanding the gestural competence of nonsigners and the ways in which this type of competence may contribute to the grammaticalization of these properties in a sign language.

Hierarchical Prosody Conversion Using Regression-Based Clustering for Emotional Speech Synthesis

This paper presents an approach to hierarchical prosody conversion for emotional speech synthesis. The pitch contour of the source speech is decomposed into a hierarchical prosodic structure consisting of sentence, prosodic word, and subsyllable levels. The pitch contour in the higher level is encoded by the discrete Legendre polynomial coefficients. The residual, the difference between the source pitch contour and the pitch contour decoded from the discrete Legendre polynomial coefficients, is then used for pitch modeling at the lower level. For prosody conversion, Gaussian mixture models (GMMs) are used for sentence- and prosodic word-level conversion. At subsyllable level, the pitch feature vectors are clustered via a proposed regression-based clustering method to generate the prosody conversion functions for selection. Linguistic and symbolic prosody features of the source speech are adopted to select the most suitable function using the classification and regression tree for prosody conversion. Three small-sized emotional parallel speech databases with happy, angry, and sad emotions, respectively, were designed and collected for training and evaluation. Objective and subjective evaluations were conducted and the comparison results to the GMM-based method for prosody conversion achieved an improved performance using the hierarchical prosodic structure and the proposed regression-based clustering method.

Modeling the Expressivity of Input Text Semantics for Chinese Text-to-Speech Synthesis in a Spoken Dialog System

This work focuses on the development of expressive text-to-speech synthesis techniques for a Chinese spoken dialog system, where the expressivity is driven by the message content. We adapt the three-dimensional pleasure-displeasure, arousal-nonarousal and dominance-submissiveness (PAD) model for describing expressivity in input text semantics. The context of our study is based on response messages generated by a spoken dialog system in the tourist information domain. We use the P (pleasure) and A (arousal) dimensions to describe expressivity at the prosodic word level based on lexical semantics. The D (dominance) dimension is used to describe expressivity at the utterance level based on dialog acts. We analyze contrastive (neutral versus expressive) speech recordings to develop a nonlinear perturbation model that incorporates the PAD values of a response message to transform neutral speech into expressive speech. Two levels of perturbations are implemented-local perturbation at the prosodic word level, as well as global perturbation at the utterance level. Perceptual experiments involving 14 subjects indicate that the proposed approach can significantly enhance expressivity in response generation for a spoken dialog system.