Word Juncture Research Articles

Word juncture characteristics in world Englishes: A research report

ABSTRACTThe subtle juncture cues in older varieties of English such as Received Pronunciation can be difficult for speakers of new English varieties to perceive. This study looks at the perception of word juncture characteristics in three varieties of English (British, Hong Kong and Singapore) among British, Hong Kong and Singaporean listeners in order to widen our understanding of English juncture characteristics in general. We find that, even though reaction time data indicates that listeners perform quickest in the variety they are most familiar with, not only are juncture differences in British English difficult for Hong Kong and Singaporean listeners to perceive, they are also the most difficult for British listeners. Juncture characteristics in Hong Kong English are the easiest to distinguish among the three varieties.

A Phonetic Investigation of Single Word versus Connected Speech Production in Children with Persisting Speech Difficulties Relating to Cleft Palate

Objective : To investigate the phonetic and phonological parameters of speech production associated with cleft palate in single words and in sentence repetition in order to explore the impact of connected speech processes, prosody, and word juncture on word production across contexts. Participants : Two boys (aged 9years 5months and 11years 0months) with persisting speech impairments related to a history of unilateral cleft lip and palate formed the main focus of the study; three typical adult male speakers provided control data. Method : Audio, video, and electropalatographic recordings were made of the participants producing single words and repeating two sets of sentences. The data were transcribed and the electropalatographic recordings were analyzed to explore lingual-palatal contact patterns across the different speech conditions. Acoustic analysis was used to further inform the perceptual analysis and to make specific durational measurements. Results : The two boys' speech production differed across the speech conditions. Both boys showed typical and atypical phonetic features in their connected speech production. One boy, although often unintelligible, resembled the adult speakers more closely prosodically and in his specific connected speech behaviors at word boundaries. The second boy produced developmentally atypical phonetic adjustments at word boundaries that appeared to promote intelligibility at the expense of naturalness. Conclusion : For older children with persisting speech impairments, it is particularly important to examine specific features of connected speech production, including word juncture and prosody. Sentence repetition data provide useful information to this end, but further investigations encompassing detailed perceptual and instrumental analysis of real conversational data are warranted.

Idiosyncratic gesture use in atypical language development, and its interaction with speech rhythm, word juncture, syntax, pragmatics and discourse: a case study

Very little is known about the use of gesture by children with developmental language disorders (DLDs). This case study of ‘Lucy’, a child aged 4;10 with a DLD, expands on what is known and in particular focuses on a type of idiosyncratic “rhythmic gesture” (RG) not previously reported. A fine-grained qualitative analysis was carried out of video recordings of Lucy in conversation with the first author. This revealed that Lucy's RG was closely integrated in complex ways with her use of other gesture types, speech rhythm, word juncture, syntax, pragmatics, discourse, visual processing and processing demands generally. Indeed, the only satisfactory way to explain it was as a partial byproduct of such interactions. These findings support the theoretical accounts of gesture which see it as just one component of a multimodal, integrated signalling system (e.g. Goldin-Meadow, S. (2000). Beyond words: The importance of gesture to researchers and learners. Child Development, 71(1), 231–239), and emergentist accounts of communication impairment which regard compensatory adaptation as integral (e.g. Perkins, M. R. (2007). Pragmatic Impairment. Cambridge: Cambridge University Press.).

Intelligibility in Children with Persisting Speech Disorders

In this paper we use detailed phonetic and phonological analysis to explore the speech production of a seven-year-old child with a persisting speech disorder, across a range of speech contexts and activities: single-word naming, sentence repetition and spontaneous conversation. The analyses reveal significant differences across the different contexts, which can be related to phonetic and phonological organisation across domains larger than the single word, and in particular to specific difficulties with word juncture. The data strongly suggest that for some children a phonological analysis of single word production is unlikely to identify key characteristics of their speech output difficulties, and that due acknowledgement in the assessment of developmental speech disorders must be given not only to the tasks and contexts within which data collection takes place, but also to the nature of phonetic and phonological organisation within the context of interaction.

Perception of allophonic cues to English word boundaries by Japanese second language learners of English

Perception of stop aspiration and glottal stop allophonic cues for word juncture in English by Japanese second language (L2) learners of English was examined, extending a study of Spanish L2 learners [Altenberg, E. P. (2005). Second Lang. Res. 21, 325-358]. Thirty Japanese listeners ranging in length of residence (LOR) in the United States (2 weeks-12 years) were tested on 42 contrasting pairs (e.g., aspiration: keeps talking vs keep stalking, glottal stop: a nice man vs an ice man, and double cues: grape in vs grey pin). Phrases were presented in randomly ordered lists and subjects responded in a two-choice identification task followed by a phrase familiarity test. The Japanese listeners performed more poorly than an American English-speaking control group (N=10), especially on aspiration pairs. Aspiration pairs were differentiated significantly less well (73% correct) by Japanese listeners than were glottal stop pairs (91% correct) and double cue pairs (94% correct); response biases predicted from relative familiarity of phrases were evident only for aspiration pairs. Performance correlated with LOR and suggested that aspiration cues take more immersion experience to learn than glottal stop cues. The patterns of errors were similar, but not identical, to Altenberg's Spanish data.

Perception of word boundaries in English speech by Japanese L2 learners of English

This study investigated the perception of aspiration and glottal stop phonetic cues for word juncture in English phrases by Japanese L2 learners of English, replicating a study of Spanish L2 learners [E. P. Altenberg, Second Language Research. 21, 325–358 (2005)]. Thirty Japanese and 10 American listeners responded to 84 phrases in randomly ordered lists containing 42 contrasting pairs, e.g., keeps talking versus keep stalking (aspiration‐cued contrasts) and a nice man versus an ice man (glottal stop‐cued contrasts) in a two‐choice task with confidence ratings. Results indicated that aspiration cues were less available to Japanese listeners (mean=73% correct) than glottal stop cues (mean=91% correct); Americans performed near ceiling on both types of contrasts (96% and 97% correct). Japanese listeners had the most difficulty perceiving aspiration‐cued contrasts containing consonant clusters at word junctures, e.g., keeps canning versus keep scanning, tops crawled versus top scrawled. JapaneseEconfidence ratings were high even when they erred, while those of native listeners better reflected response accuracy. The results support the notion that L1/L2 differences in both phonetic and phonotactic structure significantly affects perception of L2 word boundaries.

Word juncture behaviours in young children's spontaneous speech production

The present study explores word boundary behaviours in the spontaneous speech of a group of 6 preschool children. Speech collected in play settings is examined for the presence of normal and atypical connected speech behaviours, and to identify specific instances of open and close word juncture behaviours. The findings suggest that developmental and adult‐like features co‐exist in the speech of typically‐developing children and that developmental changes occur in connected speech behaviours between the ages of 2–3 years. In particular, a shift from equal numbers of close and open junctures, to a preference for close junctures is noted. Individual differences are also apparent between speakers, and these are discussed.

Perceiving non-native speech: Word segmentation

One important source of information listeners use to segment speech into discrete words is allophonic variation at word junctures. Previous research has shown that non-native speakers impose their native-language phonetic norms on their second language; as a consequence, non-native speech may (in some cases) exhibit altered patterns of allophonic variation at word junctures. We investigated the perceptual consequences of this for word segmentation by presenting native-English listeners with English word pairs produced either by six native-English speakers or six highly fluent, native-French speakers of English. The target word pairs had contrastive word juncture involving voiceless stop consonants (e.g., why pink/wipe ink; gray ties/great eyes; we cash/weak ash). The task was to identify randomized instances of each individual target word pair (as well as control pairs) by selecting one of four possible choices (e.g., why pink, wipe ink, why ink, wipe pink). Overall, listeners were more accurate in identifying target word pairs produced by the native-English speakers than by the non-native English speakers. These findings suggest that one contribution to the processing cost associated with listening to non-native speech may be the presence of altered allophonic information important for word segmentation. [Work supported by NIH/NIDCD.]

Phonotactic and Prosodic Effects on Word Segmentation in Infants

This research examines the issue of speech segmentation in 9-month-old infants. Two cues known to carry probabilistic information about word boundaries were investigated: Phonotactic regularity and prosodic pattern. The stimuli used in four head turn preference experiments were bisyllabic CVC·CVC nonwords bearing primary stress in either the first or the second syllable (strong/weak vs. weak/strong). Stimuli also differed with respect to the phonotactic nature of their cross-syllabic C·C cluster. Clusters had either a low probability of occurring at a word juncture in fluent speech and a high probability of occurring inside of words (“within-word” clusters) or a high probability of occurring at a word juncture and a low probability of occurring inside of words (“between-word” clusters). Our results show that (1) 9-month-olds are sensitive to how phonotactic sequences typically align with word boundaries, (2) altering the stress pattern of the stimuli reverses infants' preference for phonotactic cluster types, (3) the prosodic cue to segmentation is more strongly relied upon than the phonotactic cue, and (4) a preference for high-probability between-word phonotactic sequences can be obtained either by placing stress on the second syllable of the stimuli or by inserting a pause between syllables. The implications of these results are discussed in light of an integrated multiple-cue approach to speech segmentation in infancy.

On the use of inter-word context-dependent units for word juncture modeling

Word juncture coarticulation is one of the major sources of acoustic variability for initial and final word segments when spoken in fluent speech. One way to improve characterization of word pronuciations in continuous speech is to include inter-word contexts in lexical representations, similar to the way intra-word contexts are utilized. In this paper we investigate the issues related to the modeling of this set of inter-word, context-dependent units using continuous density hidden Markov models. Under such a modeling framework, it is usually required to have enough training tokens available for each unit in order reliably to estimate the parameters of the unit. Therefore, each context-dependent unit is included in the set of units to be modeled only when its frequency of occurrence in the training data exceeds a prescribed threshold. Testing such a unit selection and modeling strategy on the DARPA resource management task, it was found that the incorporation of inter-word units gave a 15–25% word error reduction compared to the base-line continuous speech recognition system using only intra-word units.

Improved acoustic modeling for large vocabulary continuous speech recognition

We report on some recent improvements to an HMM-based, continuous speech recognition system which is being developed at AT&T Bell Laboratories. These advances, which include the incorporation of inter-word, context-dependent units and an improved feature analysis, lead to a recognition system which gives a 95% word accuracy for speaker-independent recognition of the 1000-word DARPA resource management task using the standard word-pair grammar (with a perplexity of about 60). It will be shown that the incorporation of inter-word units into training results in better acoustic models of word juncture coarticulation and gives a 20% reduction in error rate. The effect of an improved set of spectral and log-energy, features is further to reduce word error rate by about 30%. Since we use a continuous density HMM to characterize each subword unit, it is simple and straightforward to add new features to the feature vector (initially a 24-element vector, consisting of 12 cepstral and 12 delta cepstral coefficients). We investigate augmenting the feature vector with 12 second difference (delta-delta) cepstral coefficients and with first (delta) and second difference (delta-delta) log energies, thereby giving a 38-element feature vector. Additional error rate reductions of 11% and 18% were achieved, respectively. With the improved acoustic modeling of subword units, the overall error rate reduction was over 42%. We also found that the spectral vectors, corresponding to the same speech unit, behave differently statistically, depending on whether they are at word boundaries or within a word. The results suggest that intra-word and inter-word units should be modeled independently, even when they appear in the same context. Using a set of subword units which included variants for intra-word and inter-word, context-dependent phones, an additional decrease of about 6–10% in word error rate resulted.

Word juncture modeling using phonological rules for HMM-based continuous speech recognition

Words uttered in isolation are pronounced differently than when they are uttered in continuous speech, a major cause being coarticulation at word junctures. Between-word context-dependent phones have been proposed to provide a more precise phonetic representation of word junctures. This technique permits one to accurately model “soft” pronunciation changes (changes in which a phone undergoes a comparatively small alteration). However, “hard” pronunciation changes (changes in which a phone is completely deleted or replaced by a different phone) are much less frequent and hence cannot be modeled adequately due to the lack of training material. To overcome this problem we use a set of phonological rules to redefine word junctures, specifying how to replace or delete the boundary phones according to the neighboring phones. No new speech units are required, thus avoiding most of the training issues. Results, which are evaluated on the 991-word speaker-independent DARPA task, show that phonological rules are effective in providing corrective capability at low computational cost.

Word Juncture in Latin Prose and Poetry

Word Juncture in Latin Prose and Poetry

Keyword and phrase spotting by use of the Harpy speech system

Two problems of a keyword spotting system that are not encountered in isolated work recognition systems are the detection of the beginning and ending of the words within the context of other words and the phenomena that occur at the word junctures. A modification of the Harpy connected speech recognition system has yielded a word and phrase spotting system that deals with these problems. This word spotting system incorporates the lexical and word juncture knowledge and uses a modified beam search technique of the Harpy system. This enables the system to not only rapidly and accurately spot single words within any context but also multiple words and/or complex phrases. Performance results of the system will be presented for various words and phrases and for different threshold values.

Locus of segmental cues for word juncture.

Listening experiments with spliced speech showed that cues for the perception of word juncture occurred always at word onset, at word offset only for /l/ and /r/, and never word medially. Spectrograms showed the cues to be bursts, aspiration, glottal stops, laryngealization, and distinct syllable-initial allophones of /l/ and /r/.

Word juncture perception as a function of word-initial phonemes

A previous study showed that allophonic variations in word-initial phonemes were the primary cues for word juncture perception. This study examined juncture perception as a function of the word-initial phonemes occurring singly and in clusters. Pairs of two-word phrases contrasting in juncture location (e.g., “he praises” and “heap raises”) and representing a wide variety of word-initial vowels, consonants and consonant clusters were recorded in sentence contexts by four talkers. The phrases were excised from the sentences and played to listeners who indicated which phrase they heard. Juncture perception was good for word-initial voiceless stops, vowels, affricates, and nasals, and fair for fricatives and voiced stops. Juncture perception was poor for liquids and semivowels, but good when they occurred in a consonant cluster. Juncture perception for clusters beginning with /s/ was better than for /s/ alone. Juncture perception was poor when the same consonant occurred before and after the juncture (e.g., “heap praises”); this suggests that segmental duration is not an important cue for juncture perception.

Study of the acoustic features of word juncture using speech analysis and synthesis

People can usually distinguish the word juncture in contrastive twins of word pairs such as play nice and plain ice. We can see many features which differ between the twins in spectrograms, but observation alone cannot tell us which features ate important to the perception of juncture. The contributions of pitch, amplitude, timing, and segmental features to juncture perception were studied as follows. Parametric representations of twins X and Y were obtained by analyzing natural speech, and the above features were exchanged between the twins in a complete factorial design. A feature f was considered important for juncture perception if exchanging f between X and Y also caused perception to be exchanged; that is, listeners heard Y when presented with X possessing feature f of Y, and vice versa. Listeners' responses indicated that all the features were important, but the importance of any particular feature or set of features depended on the juncture consonant. Features also interacted prominently for some consonants

Frequency of Occurrence of French Phonemes and Consonant Clusters

Abstract This fourth article in a series presenting the results of a computer-assisted analysis of a tape-recorded corpus of upper-class French conversations deals with the distribution and frequency of occurrence of phonemes in that language. Statistics are given for all phonemes in word-initial and word-final positions, together with grand totals for all positions, and for consonant clusters, as they occur within words and as they are created by word juncture. The numerical results are interpreted in the context of several prevalent notions about French.

Representation and Use of Acoustic-Phonetic Knowledge in the HEARSAY System

The HEARSAY speech understanding system uses acoustic-phonetic knowledge at basically four levels: lexical, syllabic, phonemic, and subsegmental. The lexicon is a list of phonemic spellings of all the words of the given language, possibly with alternates. The system recognizes possible word juncture problems from the phonemic dictionary; for example, the juncture between the words “got to” may be realized with one or two stop releases, and the system is sensitive to both possibilities. At a (roughly) syllabic level, the speech is segmented into the three classes of “voiced,” “silent,” and “fricated” (voiced or unvoiced). (In addition, a “voiced” segment is further subdivided if it contains a significant amplitude minimum between significant maxima.) This segmentation, while crude, is very reliable for many first-lever approximations. At the phonemic level, the system matches an expectation derived from the dictionary spelling to the actual utterance; each candidate word is rated as a function of the goodness of this phonemic-phonetic match. At the subsegmental level, a label is assigned to each 10-msec sample of speech. The inventory of labels is intended to correspond to the steady-state portions of the sustained phone types of English. This labeling is used to characterize each 10 msec for the further higher-level processing.