Mandarin Sentences Research Articles

Development and validation of the Mandarin disyllable recognition test

Conclusion: Given the phonetic balancing across lists and the validation with spectrally degraded speech, the present Mandarin disyllable recognition test (DRT) materials may be useful for assessing speech performance of Mandarin-speaking cochlear implant (CI) users. If combined with the previously developed sentence materials, these materials would help to establish standardized speech perception tests for Mandarin-speaking hearing-impaired (HI) and CI patients. Objectives: To develop standardized Mandarin DRT materials that can be used to evaluate the speech performance of Mandarin-speaking HI and CI patients, and to establish standardized Mandarin speech perception test materials that include both disyllables and sentences. Methods: Ten phonetically balanced Mandarin DRT lists were developed. The DRT materials were validated in 8 normal-hearing (NH) subjects listening to unprocessed speech and in 10 NH subjects listening to a 4-channel, sine-wave vocoded acoustic simulation of CI speech processing. Performance with the DRT materials was compared to that with Mandarin sentence materials previously developed by our group. Results: The distribution of vowels, consonants, and tones within each DRT list was similar to that observed across commonly used Chinese characters. There was no significant difference in disyllable word recognition across lists in both unprocessed and four-channel vocoded speech. There was a significant correlation between disyllable and sentence recognition performance.

Voice Low Tone to High Tone Ratio, Nasalance, and Nasality Ratings in Connected Speech of Native Mandarin Speakers: A Pilot Study

Voice low tone to high tone ratio (VLHR) is defined as the power ratio of a voice spectrum with a specific cut-off frequency. Previous studies have shown that there are significant correlations between VLHR and nasalance and hypernasality ratings in vowels. The correlation was investigated in this study using connected speech material. The Zoo Passage, the Rainbow Passage, the English Nasal Sentences, the Mandarin Nonnasal Sentences, and the Mandarin Nasal Sentences were used to acquire VLHRs, nasalance scores, and perceptual judgments of nasality. Each passage was recorded twice for averaging, and the cut-off frequencies from 200 Hz to 1200 Hz were used to survey for the presence of optimal correlations with VLHR. Participants : Ten native Mandarin speakers with an English learning history of over 8 years were enrolled. Main outcome measures : VLHRs, nasalance scores, and hypernasality ratings. The correlations of VLHR with nasalance (rho = .76, p < .001, Spearman rank correlation) and nasality ratings (rho = .81, p < .001) were significant using a cut-off frequency of 300 Hz for the English passages. For the Mandarin Sentences, the optimal correlations of VLHR with nasalance (rho = .83, p < .001) and nasality ratings (rho = .79, p < .001) were identified using a cut-off frequency of 500-Hz. The significant correlations of VLHR with nasalance and perceptual ratings of nasality using connected speech show that these approaches have a potential value in terms of basic and clinical application.

Modeling pitch contour of Chinese Mandarin sentences with the PENTA model

In continuous speech, the pitch contour of the same syllable may vary much due to its contextual information. The Parallel Encoding and Target Approximation (PENTA) model is applied here to Mandarin speech synthesis with a method to predict pitch contours for Chinese syllables with different contexts by combining the Classification And Regression Tree (CART) with the PENTA model to improve its prediction accuracy. CART was first used to cluster the syllables' normalized pitch contours according to the syllables contextual information and the distances between pitch contours. The average pitch contour was used to train the PENTA model with the average contour for each cluster. The initial pitch is required with the PENTA model to predict a continuous pitch contour. A Pitch Discontinuity Model (PDM) was used to predict the initial pitches at positions with voiceless consonants and prosodic boundaries. Initial tests on a Chinese four-syllable word corpus containing 2048 words were extended to tests with a continuous speech corpus containing 5445 sentences. The results are satisfactory in terms of the Root Mean Square Error (RMSE) comparing the predicted pitch contour with the original contour. This method can model pitch contours for Mandarin sentences with any text for speech synthesis.

Informational masking of speech produced by speech-like sounds without linguistic content

This study investigated whether speech-like maskers without linguistic content produce informational masking of speech. The target stimuli were nonsense Chinese Mandarin sentences. In experiment I, the masker contained harmonics the fundamental frequency (F0) of which was sinusoidally modulated and the mean F0 of which was varied. The magnitude of informational masking was evaluated by measuring the change in intelligibility (releasing effect) produced by inducing a perceived spatial separation of the target speech and masker via the precedence effect. The releasing effect was small and was only clear when the target and masker had the same mean F0, suggesting that informational masking was small. Performance with the harmonic maskers was better than with a steady speech-shaped noise (SSN) masker. In experiments II and III, the maskers were speech-like synthesized signals, alternating between segments with harmonic structure and segments composed of SSN. Performance was much worse than for experiment I, and worse than when an SSN masker was used, suggesting that substantial informational masking occurred. The similarity of the F0 contours of the target and masker had little effect. The informational masking effect was not influenced by whether or not the noise-like segments of the masker were synchronous with the unvoiced segments of the target speech.

Noise Reduction Using Wavelet Thresholding of Multitaper Estimators and Geometric Approach to Spectral Subtraction for Speech Coding Strategy

ObjectivesNoise reduction using wavelet thresholding of multitaper estimators (WTME) and geometric approach to spectral subtraction (GASS) can improve speech quality of noisy sound for speech coding strategy. This study used Perceptual Evaluation of Speech Quality (PESQ) to assess the performance of the WTME and GASS for speech coding strategy.MethodsThis study included 25 Mandarin sentences as test materials. Environmental noises including the air-conditioner, cafeteria and multi-talker were artificially added to test materials at signal to noise ratio (SNR) of -5, 0, 5, and 10 dB. HiRes 120 vocoder WTME and GASS noise reduction process were used in this study to generate sound outputs. The sound outputs were measured by the PESQ to evaluate sound quality.ResultsTwo figures and three tables were used to assess the speech quality of the sound output of the WTME and GASS.ConclusionThere is no significant difference between the overall performance of sound quality in both methods, but the geometric approach to spectral subtraction method is slightly better than the wavelet thresholding of multitaper estimators.

Development of a corpus of Mandarin sentences in babble with homogeneity optimized via psychometric evaluation

Objective: To develop a corpus of sentences in babble noise that is suitable for Mandarin-speaking children. Two experiments were conducted with specific aims of (1) developing sentence material that is grammatically and semantically within the linguistic abilities of children; and (2) improving the efficiency of the test by equalizing the relative intelligibility of individual items in sentences. Design and Study sample: Sentences were extracted from spoken material of Chinese children aged between 4 and 5 years of age. The sentences were tested for intelligibility in a four-talker babble by 96 adult native speakers of Mandarin. Psychometric functions were generated, and used for adjusting signal-to-noise ratios of individual items by varying the level of the time-locked babble to equate intelligibility of the target speech. These adjusted stimuli were tested for intelligibility using a different group of 64 adult listeners. Results: The signal-to-noise ratio for 50% correct was not different before and after adjustments (− 6.1 dB and − 6.0 dB, respectively). However, there was a significant reduction in standard deviation from 2.3 dB before adjustment to 1.1 dB after adjustment (p < 0.05). Conclusions: The experiments established a corpus of Mandarin BKB-like sentences with four-talker babble as competing noise, in which the test items’ homogeneity was optimized via psychometric evaluation (HOPE).

Lexical and syntactic representations in closely related languages: Evidence from Cantonese–Mandarin bilinguals

Bilinguals appear to have shared syntactic representations for similar constructions between languages but retain distinct representations for noncognate translation-equivalents ( Schoonbaert, Hartsuiker, & Pickering, 2007). We inquire whether bilinguals have more integrated representations of cognate translation-equivalents. To investigate this, we report two structural priming experiments in which participants heard dative sentences in Mandarin or Cantonese and described pictures using Mandarin (Experiment 1) or Cantonese (Experiment 2). We found that cognate verbs between the prime and the target led to a smaller boost than same verbs. This difference in priming could not be attributed to more phonological overlap between same verbs (i.e., identical) than between cognate verbs (i.e., similar but not identical). These results suggest that cognate verbs have separate rather than shared lemma representations across languages, even though their associated syntactic information appears to be collectively represented. Furthermore, we found an advantage for within-language priming over between-language priming. We interpreted this advantage as the result of a language node passing activation to all the lemmas linked to it. Implications for bilingual lexical and syntactic representation and processing are discussed.

Cross‐Language Priming of Word Meaning During Second Language Sentence Comprehension

The experiment investigated the benefit to second language (L2) sentence comprehension of priming word meanings with brief visual exposure to first language (L1) translation equivalents. Native English speakers learning Mandarin evaluated the validity of aurally presented Mandarin sentences. For selected words in half of the sentences there was synchronized visual exposure for 100 ms of English translation equivalents. This reduced response times relative to sentences without L1 word exposure, and exposure of translation equivalents for an early word in a sentence was instrumental in reducing comprehension errors. Results are discussed with respect to (a) the unique demands on working memory of L2 speech comprehension, (b) the meaning first and sentence location principles of attention allocation in L2 comprehension, and (c) the potential utility of cross‐language meaning activation to reduce the frequency of cognitive overload during L2 listening and thereby facilitate the acquisition of a broader range of L2 comprehension skills.

How universal is the Universal Grinder?

In this paper we investigate the interpretation of the bare noun in the Mandarin sentence in (1a) and the consequences it has for the semantic denotation of mass nouns and count nouns in the lexicon. The sentence in (1a) has what we would like to call the “wall-paper reading”: it is only felicitous to describe a situation in which a wall has been decorated with numerous little dogs. The noun gǒu ‘dog’ does not have a mass reading. Note that for the mass reading we need gǒu-rou ‘dog flesh/ meat’ (in (1b)). Note also, that a bare mass noun like shuǐ ‘water’ is perfect in these sentences (see (1c)).

Frequency modulated harmonics induces informational masking of Chinese speech

Informational masking (IM) occurs when target speech is co-presented with competing speech. It is important to know what acoustic characteristics of a speech masker are essential to induce IM. This study investigated whether frequency modulation (FM) of speech is associated with IM. Nonsense Chinese Mandarin sentences were spoken by a young female as target stimuli and maskers were tone harmonics whose fundamental frequency (F0) was modulated sinusoidally and number of harmonic streams was varied. The magnitude of IM was evaluated by measuring the precedence-induced releasing effect of perceived spatial separation on target speech. The results show that the releasing effect reached the highest level when F0 became 252 Hz, which was the mean pitch of target speech. In addition, the releasing effect was significantly higher when the masker contained two harmonic streams than when one stream. These results suggest that because frequency-modulated harmonics can be used to inducing IM of speech, the FM component of speech is an important acoustic factor underlying IM. [Supported by the NSFC 60435010; 60535030; 60605016; 30670704.]

Development of the Mandarin Hearing in Noise Test (MHINT)

To develop two versions of the Mandarin Hearing In Noise Test (MHINT). These tests are adaptive tests that measure the reception threshold for sentences (RTSs) in quiet and in noise. The RTS is the presentation level at which half the sentences are correctly recognized. Four studies were undertaken to (1) develop sentence materials, (2) equalize sentence difficulty, (3) create phonemically balanced sentence lists; and (4) evaluate within-list response variability, inter-list reliability, and produce normative data. A total of 137 native Mandarin (Putonghua) speaking subjects in Mainland China and 89 native Mandarin speakers in Taiwan participated. They had normal hearing thresholds at 25 dB HL or better. RTSs were measured under four headphone test conditions: Quiet, and in noise with noise originating from the 0 degree (Noise Front), 90 degrees to the right (Noise Right), and 90 degrees to the left (Noise Left). The speech originated from the front (0 degree) in all conditions. The noise level was fixed at 65 dBA, and the speech was varied adaptively to find the RTS. Two versions of the test materials, consisting of 24, 20-sentence lists each in Mandarin spoken in the Mainland (the MHINT-M) and the dialect of Mandarin spoken in Taiwan (the MHINT-T), were created from two sets of 240 sentences containing 10 syllables per sentence. The mean Quiet RTS was 14.7 dBA, using the MHINT-M, and 19.4 dBA, using the MHINT-T. Using the MHINT-M, the mean RTS for Noise Front was -4.3 dB signal-to-noise ratio (SNR), -11.7 dB SNR for Noise Right, and -11.7 dB SNR for Noise Left. Using the MHINT-T, the Noise Front RTS was -4.0 dB SNR, -10.7 dB SNR for Noise Right, and -11.0 dB SNR for Noise Left. Results in noise are slightly better than those seen for the English HINT norms. Response variability within list was low, and inter-list reliability was high, indicating that consistent results can be obtained using any list. Confidence intervals are reported. The two versions of the MHINT are the first standardized Mandarin sentence speech intelligibility tests. Similar to other language versions of the HINT, the MHINT was developed using the same rationale as the English HINT, allowing norm-referenced results for the MHINT to be compared directly with results in other languages. The MHINT would benefit from further evaluation of its validity.

New and contrastive focus in Taiwan Mandarin

This study investigated the influence of information structure on surface acoustical parameters in Taiwan Mandarin. There were two experiments, i.e., GN and GC experiment. In GN experiment, the noun phrases placed in initial (NP1) or medial (NP2) position in a Mandarin sentence were elicited through a game. Subjects answered experimenter’s questions, e.g., ‘‘What is on pink truck’s right side?’’ according to colored pictures in a 4 by 4 matrix. New or given information was placed on either NP1 or NP2. In GC experiment, subjects corrected or confirmed expeirmenter’s description, such as Pink truck is yellow IDs right side, by placing either contrastive information on the color term and given information on noun (cg), given information on the color term and contrastive information on noun (gc), given information on both color term and noun (gg), or contrastive information on both color term and noun (cc). Results showed that the F0 range expansion and duration lengthening were observed on syllables carrying new and contrastive information. Moreover, duration lengthening is more pronunced than F0 range expansion. Furthermore, syllables with contrastive information were lengthened more than syllables with new information.

A cross-language fMRI study of sentence-level prosody in Mandarin

Speech prosody engages a large-scale spatially distributed network in both hemispheres (Friederici & Alter, 2004; Hickok & Poeppel, 2004). It may be conceptualized as a mosaic of multiple local asymmetries that allows for the possibility that different regions may be differentially weighted in laterality depending on language-, modality-, and task-related features (Scott & Wise, 2004). Cross-language comparisons of speech prosody have demonstrated that hemispheric laterality effects vary depending on the role of specific brain region at different stages of processing (Gandour et al., 2004). Mandarin Chinese is a tone language that offers unique opportunities to address issues of functional asymmetry that involve prosodic contrasts based primarily on variations in pitch. In addition to its four lexical tones, pitch variations are also exploited to signal post-lexical, prosodic phenomena at the sentence level: narrow focus and intonation. Using functional magnetic resonance imaging (fMRI), this study investigates the perception of sentence focus and intonation in Mandarin in an affectively neutral context. Whereas our earlier work focused on word, syllable, and phonemic units of prosody, or comparisons of wordto sentence-level units, this study directly compares two sentence-level units. Sentences were designed with two sentence types (statement, question) in combination with two locations of sentence focus or emphatic stress (initial, final). For example (bold= focused word; superscript numbers=Chinese tones) bi ge cao. ‘Bi mows lawns.’, statement/ initial focus; kou du shu. ‘Kou reads books.’, statement/final focus; song da qiu? ‘Song plays golf?’, question/initial focus; dai hua tu? ‘Dai draws maps?’, question/final focus. Chinese and English listeners were asked to selectively attend to either focus (F) or intonation (I) in paired three-word sentences, and make speeded-response discrimination judgments. Scanning was performed on a 1.5T Signa GE LX Horizon scanner. Functional data were acquired using a gradient-echo EPI pulse sequence. Image analysis was conducted using the AFNI software package (Cox, 1996). Nine anatomically constrained, 6-mm radius spherical regions of interest (ROI) were chosen that have been implicated in previous functional neuroimaging studies of phonological processing, speech perception, semantic processing, attention, and working memory. ROIs were symmetric in nonoverlapping frontal, temporal, and parietal regions of both hemispheres. For each ROI, mean z scores were calculated for each task (F, I) and hemisphere (LH, RH) for every subject. These z scores within each ROI were analyzed using ANOVAs to compare activation between tasks (F, I), hemispheres (LH, RH), and groups (Chinese, English). Results showed that both language groups share right-sided activity in the mid portion of the middle frontal gyrus (BA 46/9) across tasks, and bilateral activity in the dorsal aspect of inferior parietal cortex (BA 40/7) in a direct task comparison (focus > intonation). The Chinese group, on the other hand, exhibited stronger activity than the English group across tasks in the left-sided anterolateral ventral aspect of the supramarginal gyrus (BA 40) and posterior aspect of the middle temporal gyrus (BA 21/20/37). Language-dependent task effects (focus > intonation) were observed in the left posterior middle temporal gyrus (BA 21/20/37) for the Chinese group. The major findings of this study demonstrate that brain activity in response to Mandarin sentence focus and intonation depends on languageand task-related features. All regions in the frontal, temporal, and parietal lobes that are lateralized to the LH in response to one or both tasks are found in the Chinese group only. Conversely, one regions in the frontal lobe that is lateralized to the RH is found in both language groups. No region shows a leftward asymmetry in the English group. Instead of viewing hemispheric roles as singular in nature, we argue that there are multiple levels of specialization that are associated with different stages of processing (e.g., acoustic, linguistic). Such findings support the emerging view that speech prosody perception involves a dynamic interplay among widely distributed regions not only within a single hemisphere but also between the two hemispheres (Friederici & Alter, 2004). Although language experience can shape the internal representation of an external auditory signal, it is not a question of whether any single factor by itself—language experience, stimulus parameter, stimulus function, task demands—is the driving force behind hemispheric laterality of speech prosody. All of the above are relevant but may vary in their individual weight depending on experimental design and, as a consequence, the degree of involvement of the LH. Brain and Language 95 (2005) 54–55

Mandarin Speech Perception in Nucleus CI 24 Implantees Using MAPs Based on Neural Response Telemetry

The purpose of this study is to compare speech perception performance in Mandarin-speaking Nucleus CI24 implantee using standard behavior MAPs and NRT-based MAPs. Eight Nucleus CI24 users (5 years and older) participated in the study. They all fulfilled the following criteria: (1) behavioral MAP and NRT thresholds can be reliably obtained; (2) had more than 18 functioning electrodes; (3) had at least 6 months experience using CI. All subjects received speech evaluation under three different MAPs: a traditional behavioral MAP, a MAP predicted from the NRT thresholds of the E22 (electrode 22), E19, E15, E11, E8, E5, E1 and a combined MAP based on the information of NRT thresholds and behavioral threshold/comfortable levels of the E11. The speech evaluation package included word recognition test in quiet, in noise, and a Mandarin sentence test in quiet. Results showed that three MAPs are similar in some subjects, but different in other subjects. Compared to the NRT MAPs, the combined MAPs are more similar to the behavioral MAPs. There was no significant difference in the mean score of the word recognition test in quiet, in noise and sentence test under these three MAP conditions. In conclusion, although the behavioral MAPs and the NRT-based MAPs are not identical, the speech performance of Mandarin-speaking CI24 implantee using MAPs predicted from NRT thresholds appeared to be no worse than the traditional behavioral MAPs. Therefore, in certain cases that behavioral MAPs are difficult to obtain (such as in very young or multiple handicapped children), NRT-based MAPs may serve reliably as an initial estimation.

Recognizing focus in noise filled sentences

This study is designed to help identify the intrinsic constituents of focus. Twelve four-word Mandarin sentences were recorded by a native speaker five times, each time either with focus on one of the words, or without any focus. Then, one, two or three words in each sentence produced by the speaker were replaced by pink noise. The noise-filled sentences were presented to subjects along with the text. The subjects’ task was to determine if the sentence had a focus, and if yes, on which word. Ten native Mandarin speakers participated as subjects. Their performance was compared across noise replacement conditions. It was found that, when both on-focus and post-focus words were present, focus could be recognized consistently. When only the focused word was present, focus could be recognized fairly well unless the focus position was sentence final, in which case it was not very distinct from no focus. When post-focus word(s) was(were) left intact while focused words were replaced by noise, focus could still be detected successfully, but its exact localization was sometimes judged wrong. These results seem to support the dual-component hypothesis about focus. Further implications of the findings will be discussed.

Processing of visually presented sentences in Mandarin and English studied with fMRI.

Comprehension of visually presented sentences in fluent bilinguals was studied with functional magnetic resonance imaging (fMRI) using a set of conceptually similar sentences in two orthographically and phonologically distinct languages, Mandarin and English. Responses were monitored during scanning. Sentence comprehension in each language was compared to fixation in nine subjects and Tamil-like pseudo-word strings in five subjects. Spatially congruent activations in the prefrontal, temporal, and superior parietal regions and in the anterior supplementary motor area were observed for both languages and in both experiments at the individual and group levels of analysis. Proficient bilinguals exposed to both languages early in life utilize common neuroanatomical regions during the conceptual and syntactic processing of written language irrespective of their differences in surface features.

Effects of tone and focus on the formation and alignment of f 0contours

The present study examines how lexical tone and focus contribute to the formation and alignment of f 0contours in speech. This was done through an investigation of f 0contour formation in short Mandarin sentences. These sentences all consisted of five syllables with varying tones on the middle three syllables. The sentences were produced by eight Mandarin speakers with four different focus patterns: focus on the first, second, or last word, or with no narrow focus. The f 0patterns of these sentences were examined through point-by-point f 0tracing, graphical comparison of averaged f 0contours, f 0-contour-syllable alignment analysis, and analysis of maximum, minimum f 0, and slope of f 0contours. The results indicate that (a) while the lexical tone of a syllable is the most important determining factor for the local f 0contour of the syllable, focus extensively modulates the global shape of the f 0curve, which in turn affects the height and even the shape of local contours; (b) the tones of adjacent syllables also extensively influence both the shape and height of the f 0contour of a syllable, with the preceding tone exerting more influence than the following tone; (c) despite extensive variations in shape and height, the f 0contour of a tone remains closely aligned with the associated syllable; and (d) both focus and tonal interaction may generate substantial f 0decline over the course of an utterance. These findings seem to be able to reduce the unpredictability in the formation and alignment of f 0contours in speech.

Units in Mandarin Conversation: Prosody, Discourse and Grammar

This book provides a new way of studying grammar. The basic thrust of the book is to investigate grammar based on a prosodic unit, the intonation unit (IU), in spontaneous speech. The author challenges the dominant practice in the study of syntax, which has been to focus on the unit of the artificially constructed sentence. The book shows that some basic notions developed from sentence-level data often do not account well for speech data. For example, in many versions of syntactic theory, the basic syntactic structure of any sentence is assumed to comprise both an NP and a VP (with variations in terminology). However the author shows that a Mandarin sentence in spoken discourse can consist of a lone NP or a transitive verbal expression without any explicit argument (which is not due to anaphora). Although the book concerns Mandarin discourse and grammar, it will be of interest to students of a wide range of fields, including discourse analysis, syntax, conversation analysis, prosodic studies, and typological studies.

The synthesis rules in a Chinese text-to-speech system

The synthesis rules developed for a successfully implemented Chinese text-to-speech system are described in detail. The design approach is based on a syllable concatenation that is rooted in the special characteristics of the Chinese language. Special attention given to the lexical tones and other prosodic rules, such as concatenation rules, sandhi rules, stress rules, intonation patterns, syllable duration rules, pause insertion rules, and energy modification rules. The rules are derived from the acoustic properties of Mandarin Chinese and therefore are useful not only in designing other Chinese text-to-speech systems, but also in understanding the characteristics of Mandarin sentences and processing Mandarin speech signals for other purposes such as segmentation or recognition. >

Intonation Variation in a Mandarin Sentence for Three Expressions: Interrogative, Exclamatory and Declarative

Abstract This study investigated two aspects of Mandarin Chinese: (1) the acoustic correlates of intonations of a sentence for three expressions, declarative, interrogative and exclamatory, and (2) the influence of intonation on the four tones in the sentence-final position. A corpus was prepared and 6 speakers recorded the corpus. Acoustical analysis was made for the 1,824 test items. Results show that the duration of syllable nuclei in a sentence varies according to word position rather than intonation except that in the sentence-final position. Both intonation and word position have influenced the fundamental frequency, tone range, and peak amplitude of syllable nuclei. The duration, fundamental frequency and peak amplitude of the four tones in sentence-final position are modified by intonation. However, the basic characteristic of the shape of the four tones remains very close to that of tones in isolated words.