Phonemic Similarity Research Articles

64. A Computational Analysis of Verbal Fluency in Schizophrenia

Abstract Background: Verbal fluency tasks are widely used to assess language organization and executive functioning. Meta-analyses of verbal fluency performance in individuals with schizophrenia (SZ) indicate greater impairment on semantic (SF) than phonemic fluency (PF). Because SF depends more on efficient word retrieval within subcategories (clustering) and PF more on shifting between subcategories (switching), fluency impairment has been linked to clustering difficulties in SZ. However, most scoring methods rely on subjective judgments to determine clusters and switches, which reduces reliability. Here, we test a novel computational method to objectively capture clustering, switching, and overall retrieval organization from fluency performance in SZ and demographically matched healthy controls (HC). Methods: PF and SF (“animals”) tasks were administered to SZ and HC. Responses were transcribed from audio recordings into text files for computational analysis using VFClust (Ryan et al, 2013). VFClust utilizes latent semantic analysis (LSA) to quantify the semantic similarity of ordered word pairs as well as a combination of the Cargenie Mellon Phonemic Dictionary (CMUDict) with the Levenshtein distance method to determine phonemic similarity of ordered word pairs. Similarity thresholds were established and applied to these pairwise similarity scores (PSS) to determine cluster or chain inclusion. Clusters included words in which every word pair in that cluster had a PSS above threshold. Chains consisted of words in which PSS of adjacent words were above threshold. Switches were counted when PSS were below threshold, signifying a shift from one category to another. Output metrics for each individual included total words produced, mean size of chains and clusters, number of chains, clusters, and switches, and mean PSS across all ordered word pairs for SF and PF. Results: SZ were impaired compared to HC on SF but not PF. SF deficit was driven by a lower number of chains and clusters produced by SZ compared with HC. Group differences in SF were not driven by any other measures. Negative symptoms were inversely correlated with the total number of words produced and the switch count in SF. Conclusion: Objective, computationally derived fluency metrics captured group differences observed in SF in SZ. Interestingly, the lack of difference in cluster size questions SZ deficits in generating sizable clusters. Instead, limitations in the ability to retrieve a variety of semantically associated words (multiple clusters) seem to underlie performance deficits. Intact switching despite lower number of clusters and chains could indicate increased activation between (conventionally) unrelated words. While altered, semantic relatedness could cascade to thought disorder and abnormal interpersonal interactions, in this study, we only observed a significant negative association with negative symptoms.

Dynamic Encoding of Acoustic Features in Neural Responses to Continuous Speech.

Humans are unique in their ability to communicate using spoken language. However, it remains unclear how the speech signal is transformed and represented in the brain at different stages of the auditory pathway. In this study, we characterized electroencephalography responses to continuous speech by obtaining the time-locked responses to phoneme instances (phoneme-related potential). We showed that responses to different phoneme categories are organized by phonetic features. We found that each instance of a phoneme in continuous speech produces multiple distinguishable neural responses occurring as early as 50 ms and as late as 400 ms after the phoneme onset. Comparing the patterns of phoneme similarity in the neural responses and the acoustic signals confirms a repetitive appearance of acoustic distinctions of phonemes in the neural data. Analysis of the phonetic and speaker information in neural activations revealed that different time intervals jointly encode the acoustic similarity of both phonetic and speaker categories. These findings provide evidence for a dynamic neural transformation of low-level speech features as they propagate along the auditory pathway, and form an empirical framework to study the representational changes in learning, attention, and speech disorders.SIGNIFICANCE STATEMENT We characterized the properties of evoked neural responses to phoneme instances in continuous speech. We show that each instance of a phoneme in continuous speech produces several observable neural responses at different times occurring as early as 50 ms and as late as 400 ms after the phoneme onset. Each temporal event explicitly encodes the acoustic similarity of phonemes, and linguistic and nonlinguistic information are best represented at different time intervals. Finally, we show a joint encoding of phonetic and speaker information, where the neural representation of speakers is dependent on phoneme category. These findings provide compelling new evidence for dynamic processing of speech sounds in the auditory pathway.

Facilitation and interference of phoneme repetition and phoneme similarity in speech production

ABSTRACTSpeakers know that it is hard to produce tongue-twister-like utterances involving repeated and similar phonemes (“She sells socks in a small shoe shop”). While utterances involving phonologically similar phonemes consistently give rise to speech errors, repeated phonemes have been shown to interfere as well as to facilitate production. Here we try to shed light on this paradox by investigating brain processes underlying planning utterances involving similar and repeated phonemes (tongue-twister-like) and utterances with only repeated phonemes elicited with picture naming during electroencephalographic/event-related potentials (EEG/ERP) recording. Tongue-twister-like utterances increased error rates while repeated onset phonemes decreased reaction times. Both types of utterances modulated ERPs after 370 ms following picture onset by shortening a period of stable electrophysiological pattern. Another microstate was lengthened for tongue-twister-like utterances only, wiping out the timing difference of the previous microstate. These results indicate priming of repeated phonemes and cost of encoding similar phonemes, both phenomena co-occurring in tongue-twisters.

Vocabulary optimization process using similar phoneme recognition and feature extraction

In processing voice with environment noise, the noise must be eliminated to improve the vocabulary recognition rate. In this process, noise elimination and feature extraction for model-estimate tec...

Variation in the lexical distribution and implementation of phonetically similar phonemes in Catalan

In some Romance languages with two pairs of mid vowel phonemes, it is acknowledged that these contrasts are somewhat unstable. We analyze the distribution and realization of the anterior and posterior mid vowels in Catalan to test claims (mostly based on anecdotal evidence) that these contrasts exhibit inter- and intraspeaker variability. Participants produced target words containing stressed mid vowels and, later, judged vowel height (/e/ vs. /ɛ/; /o/ vs. /ɔ/) in the same words. The results indicate that, even intradialectally, the distribution of mid vowels is somewhat variable, with speakers showing only moderate agreement in the distribution of phonemic vowels. In addition, speakers are not always consistent in their realization of mid vowels when they produce the same word (probably indicating weak phonolexical representations). Interspeaker variation was also observed in the phonetic implementation of the contrasts. The results indicate that the Catalan mid vowel contrasts, like those in other Romance languages, are weaker and less stable than other phonological oppositions.

Swahili Loanwords in Gorwaa and Iraqw: Phonological and Morphological Observations

When Swahili words are borrowed into Gorwaa and Iraqw, they are typically nativized according to a series of observable patterns. This work offers an account of how Swahili borrowings are integrated phonologically and morphologically into these languages. Phonologically, non-native phonemes are typically replaced by native sounds, which are featurally similar to the original Swahili phoneme, and where no such similar phoneme exists, the non-native sound is tolerated. In order to repair words whose etymon begins with a syllabic nasal, either a vowel is inserted to break up the cluster, or the initial nasal is deleted. Vowels are very often lengthened on the stressed (penultimate) syllable. Where Swahili words do not violate any phonological rule, they are adopted without change. Morphologically, borrowed verbs are assigned a verbal derivational ending to identify the forms as verbs. The gender system of Swahili differs markedly from Gorwaa and Iraqw, and, as such, every borrowed noun must be assigned a gender. Typically (but not always), this is done according to the phonetic form of the noun. Nouns ending in round vowels are assigned masculine gender and nouns ending in the front vowel are assigned feminine gender. Pluralizing forms whose base is singular, or singularizing forms whose base is plural are more complicated, with a series of weakly productive rules operating to assign number suffixes to base forms. This results in a high number of irregular forms.

Cross-linguistic perceptual learning in advanced second language listeners

An issue of long-standing interest is whether phones, which correspond to each other phonologically but differ phonetically across two languages, are distinct in the minds of bilinguals and second language learners. Here, we ask whether listeners maintain separate representations for phonetically highly similar phonemes common to two languages even though this might seem inefficient. One way to address this question is to induce a change in the representation of phonemes in one language, and to test whether the other language is also affected. We carried out two perceptual learning studies, one with intermediate to advanced L1 English learners of German and one with intermediate to advanced L1 German learners of English, to probe the relationship between listeners’ representations for the /f-s/ contrast in English and in German. We found cross-linguistic perceptual learning effects extending from English to German in both studies, though the effect sizes for perceptual learning in English and German differed. Based on these findings, we propose that phonetically highly similar phonemes common to listeners’ L1 and L2 have separate yet interconnected mental representations for speech perception.

Modeling the role of distributional information in children’s use of phonemic contrasts

Abstract Between the first and the second year of life, children improve in their ability to use phonemic contrasts when learning label–object pairings. This improvement may be related to children’s experience with the distribution of phonemes across lexical forms. Because phonemes typically occur in different lexical frames (e.g., /d/ and /t/ in “doggy” and “teddy” rather than “doggy” and “toggy”), familiarity with words makes similar phonemes more distinct through acquired distinctiveness. In a series of simulations, we demonstrate that English input has the distributional characteristics necessary to facilitate use of phonemic contrasts as a function of increasing familiarity with the lexicon. Further, these simulations support a novel prediction: that less common phonemes should take longer to be used productively. We tested this prediction with children between 18 and 25 months, and found that the relatively infrequent /s/ and /z/ contrast takes longer to emerge than frequent contrasts such as /b/–/d/ or /d/–/t/.

The relationship between speech perception and production: Evidence from children with speech production errors

Children with speech sound disorders (SSD), for reasons that are not well understood, fail to produce age-appropriate speech production targets in isolation, words, and/or functional conversation. One line of research suggests these children exhibit underlying perceptual deficits for acoustic cues that distinguish similar phonemes, exhibiting particular deficits for those phonemes produced in error (e.g., /r/, /l/) (Monnin and Huntington, 1974; Shuster, 1998), although other studies have reported mixed results in children with SSD (Locke, 1980; Rvachew and Jamieson, 1989). In the current study, children with SSD and typically developing children were presented with a perceptual discrimination task involving a phoneme produced in error by the SSD children (i.e., /r/). Findings revealed two separate subsets of children with SSD. The first exhibited poor discrimination of /r/ while a second exhibited exceptionally good discrimination of the same contrast. Given that all children with SSD exhibited distorted /r/ productions, a follow-up acoustic analysis is being conducted to identify specific acoustic characteristics of the distorted /r/s in the two groups (i.e., the poor /r/ discriminators versus the excellent /r/ discriminators). The pattern of results relating the acoustical details of /r/ production in these two populations should contribute to a more refined model of speech sound disorders.

AKRONIM YANG BERFONOTAKTIK TIDAK LAZIM DALAM BAHASA INDONESIA

This acronym research describes some uncommon phonotactic acronyms in Indonesian. Three indicators were used as guidance: (1) the list of phonotactic phonemes, (2) the common closed consonants, (3) the number of common syllables in Indonesian. The data were taken from national news papers from January 2012 thorugh March 2013. The results indicated: (1) uncommon phonotactic acronyms were usually found in the phoneme list of /md/ for Gakumdu, /nm/ for menmud, the /pk/ for Apkasi, and /pm/ for ipmi. (2) The acronyms of two identical sequenced consonants were found in the consonants /pp/ as in Bappepti, /ss/ in Kopassus, /mm/ in jamman, and /tt/ in Unpatti. (3) The acronyms of double consonant were found in the consonants /dh/ as in pusbadhi, /ngg/ in unpatti, and /nd/ in unand. The acronyms containing consonanst that were not used as word-closing acronyms were the consonants /g/ and /d/. The closing word consonant /c/ was not found in the data. Besides, acronyms using closed word consonants were found in the consonants /z/ and /v/. In borrowed words the consonants that were not found were /z/ and /v/ as word-closing acronyms, such as in baznas, pemprov. The repetition of similar phonemes in some syllables was found in riskesdas. The number of syllables commonly found in Indonesian acronyms consisted of four or more words, for example, babinkumnas, ditreskrimsus.

Lexical factors and cerebral regions influencing verbal fluency performance in MCI

ObjectiveTo evaluate assumptions regarding semantic (noun), verb, and letter fluency in mild cognitive impairment (MCI) and Alzheimer disease (AD) using novel techniques for measuring word similarity in fluency lists and a region of interest (ROI) analysis of gray matter correlates. MethodFifty-eight individuals with normal cognition (NC, n=25), MCI (n=23), or AD (n=10) underwent neuropsychological tests, including 10 verbal fluency tasks (three letter tasks [F, A, S], six noun categories [animals, water creatures, fruits and vegetables, tools, vehicles, boats], and verbs). All pairs of words generated by each participant on each task were compared in terms of semantic (meaning), orthographic (spelling), and phonemic (pronunciation) similarity. We used mixed-effects logistic regression to determine which lexical factors were predictive of word adjacency within the lists. Associations between each fluency raw score and gray matter volumes in sixteen ROIs were identified by means of multiple linear regression. We evaluated causal models for both types of analyses to specify the contributions of diagnosis and various mediator variables to the outcomes of word adjacency and fluency raw score. ResultsSemantic similarity between words emerged as the strongest predictor of word adjacency for all fluency tasks, including the letter fluency tasks. Semantic similarity mediated the effect of cognitive impairment on word adjacency only for three fluency tasks employing a biological cue. Orthographic similarity was predictive of word adjacency for the A and S tasks, while phonemic similarity was predictive only for the S task and one semantic task (vehicles). The ROI analysis revealed different patterns of correlations among the various fluency tasks, with the most common associations in the right lower temporal and bilateral dorsal frontal regions. Following correction with gray matter volumes from the opposite hemisphere, significant associations persisted for animals, vehicles, and a composite nouns score in the left inferior frontal gyrus, but for letter A, letter S, and a composite FAS score in the right inferior frontal gyrus. These regressions also revealed a lateralized association of the left subcortical nuclei with all letter fluency scores and fruits and vegetables fluency, and an association of the right lower temporal ROI with letter A, FAS, and verb fluency. Gray matter volume in several bihemispheric ROIs (left dorsal frontal, right lower temporal, right occipital, and bilateral mesial temporal) mediated the relationship between cognitive impairment and fluency for fruits and vegetables. Gray matter volume in the right lower temporal ROI mediated the relationship between cognitive impairment and five fluency raw scores (animals, fruits and vegetables, tools, verbs, and the composite nouns score). ConclusionSemantic memory exerts the strongest influence on word adjacency in letter fluency as well as semantic verbal fluency tasks. Orthography is a stronger influence than pronunciation. All types of fluency task raw scores (letter, noun, and verb) correlate with cerebral regions known to support verbal or nonverbal semantic memory. The findings emphasize the contribution of right hemisphere regions to fluency task performance, particularly for verb and letter fluency. The relationship between diagnosis and semantic fluency performance is mediated by semantic similarity of words and by gray matter volume in the right lower temporal region.

Evidence of semantic clustering in letter-cued word retrieval

Letter-cued word fluency is conceptualized as a phonemically guided word retrieval process. Accordingly, word clusters typically are defined solely by their phonemic similarity. We investigated semantic clustering in two letter-cued (P and S) word fluency task performances by 315 healthy adults, each for 1 min. Singular value decomposition (SVD) and generalized topological overlap measure (GTOM) were applied to verbal outputs to conservatively extract clusters of high-frequency words. The results generally confirmed phonemic clustering. However, we also found considerable semantic/associative clusters of words (e.g., pen, pencil, and paper), and some words showed both phonemic and semantic associations within a single cluster (e.g., pair, pear, peach). We conclude that letter-cued fluency is not necessarily a purely phonemic word retrieval process. Strong automatic semantic activation mechanisms play an important role in letter-cued lexical retrieval. Theoretical conceptualizations of the word retrieval process with phonemic cues may also need to be reexamined in light of these analyses.

Sequence Complexity Effects on Speech Production in Healthy Speakers and Speakers with Hypokinetic or Ataxic Dysarthria

The present study investigated the effects of sequence complexity, defined in terms of phonemic similarity and phonotoactic probability, on the timing and accuracy of serial ordering for speech production in healthy speakers and speakers with either hypokinetic or ataxic dysarthria. Sequences were comprised of strings of consonant-vowel (CV) syllables with each syllable containing the same vowel, /a/, paired with a different consonant. High complexity sequences contained phonemically similar consonants, and sounds and syllables that had low phonotactic probabilities; low complexity sequences contained phonemically dissimilar consonants and high probability sounds and syllables. Sequence complexity effects were evaluated by analyzing speech error rates and within-syllable vowel and pause durations. This analysis revealed that speech error rates were significantly higher and speech duration measures were significantly longer during production of high complexity sequences than during production of low complexity sequences. Although speakers with dysarthria produced longer overall speech durations than healthy speakers, the effects of sequence complexity on error rates and speech durations were comparable across all groups. These findings indicate that the duration and accuracy of processes for selecting items in a speech sequence is influenced by their phonemic similarity and/or phonotactic probability. Moreover, this robust complexity effect is present even in speakers with damage to subcortical circuits involved in serial control for speech.

Towards a theory of multimorphemic word production: The heterogeneity of processing hypothesis

Theories of spoken production have not yet addressed the post-lexical processing of multimorphemic words, that is, how a multimorphemic word's phonological form is prepared for production. This paper reviews what is known about how multimorphemic words are represented in production at lexical and post-lexical stages as well as the influence that lexical properties have on post-lexical processes. A proposal linking these facts together is presented which predicts that post-lexical processes (1) should be weaker when acting across morpheme bundaries and (2) should be influenced by the lexical properties of each morpheme. Post-lexical processing is thus predicted to vary, or be “heterogeneous”, across a multimorphemic word. Phoneme competition (as indexed by inhibitory effects of phoneme similarity) is compared within and across morphemes in three analyses of oral reading latencies. Competition is found to be weaker across morpheme boundaries, providing support for heterogeneity.

Fundamental frequency as cue to intonation: Focus on Ika Igbo and English rising intonation patterns

This paper shows that fundamental frequency, Fo, can be a cue to type of intonation. The work centers on three main intonation patterns in Ika Igbo and English. Ika Igbo is a language that manifests intonation in addition to lexical tone. These intonation patterns are Low Rise (LR), High Rise (HR) and Fall Rise (FR). The Fos of these intonation patterns were analyzed acoustically in utterances with similar phonemes and tunes in both languages. Eighteen utterances were used for the study. The analyses show that the Fos of LR and FR intonation were generally lower than those of HR. Hence it can be concluded that high intonation has high Fo while low intonation has low Fo. It can therefore be concluded that Fo is a cue to type of intonation.

Speech Recognition Error Compensation using MFCC and LPC Feature Extraction Method

Speech recognition system is input of inaccurate vocabulary by feature extraction case of recognition by appear result of unrecognized or similar phoneme recognized. Therefore, in this paper, we propose a speech recognition error correction method using phoneme similarity rate and reliability measures based on the characteristics of the phonemes. Phonemes similarity rate was phoneme of learning model obtained used MFCC and LPC feature extraction method, measured with reliability rate. Minimize the error to be unrecognized by measuring the rate of similar phonemes and reliability. Turned out to error speech in the process of speech recognition was error compensation performed. In this paper, the result of applying the proposed system showed a recognition rate of 98.3%, error compensation rate 95.5% in the speech recognition.

Speech Recognition Optimization Learning Model using HMM Feature Extraction In the Bhattacharyya Algorithm

Abstract Speech recognition system is shall be composed model of learning from the inaccurate input speech. Similar phoneme models to recognize, because it leads to the recognition rate decreases. Therefore, in this paper, we propose a method of speech recognition optimal learning model configuration using the Bhattacharyya algorithm. Based on feature of the phonemes, HMM feature extraction method was used for the phonemes in the training data. Similar learning model was recognized as a model of exact learning using the Bhattacharyya algorithm. Optimal learning model configuration using the Bhattacharyya algorithm. Recognition performance was evaluated. In this paper, the result of applying the proposed system showed a recognition rate of 98.7% in the speech recognition. Key Words : Speech Recognition, HMM, Feature Extraction, Bhattacharyya Algorithm, Optimization Learning Model Configuration * 이 논문은 2013년도 가천대학교 교내연구비 지원에 의한 결과임.(GCU-2013-R138)Received 14 May 2013, Revised 12 June 2013Accepted 20 June 2013Corresponding Author: Sangyeob Oh(The University of Gachon) Email: syoh1234@gmail.comⒸ The Society of Digital Policy & Management. All rights reserved. This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.otg/licenses/by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work ISSN: 1738-1916 is properly cited.

Understanding native Russian listeners’ errors on an English word recognition test: Model-based analysis of phoneme confusion

Objective: Word recognition is a basic component in a comprehensive hearing evaluation, but data are lacking for listeners speaking two languages. This study obtained such data for Russian natives in the US and analysed the data using the perceptual assimilation model (PAM) and speech learning model (SLM). Design: Listeners were randomly presented 200 NU-6 words in quiet. Listeners responded verbally and in writing. Performance was scored on words and phonemes (word-initial consonants, vowels, and word-final consonants). Study sample: Seven normal-hearing, adult monolingual English natives (NM), 16 English-dominant (ED), and 15 Russian-dominant (RD) Russian natives participated. ED and RD listeners differed significantly in their language background. Results: Consistent with the SLM, NM outperformed ED listeners and ED outperformed RD listeners, whether responses were scored on words or phonemes. NM and ED listeners shared similar phoneme error patterns, whereas RD listeners’ errors had unique patterns that could be largely understood via the PAM. RD listeners had particular difficulty differentiating vowel contrasts /i-I/, /æ-ε/, and /ɑ-Λ/, word-initial consonant contrasts /p-h/ and /b-f/, and word-final contrasts /f-v/. Conclusions: Both first-language phonology and second-language learning history affect word and phoneme recognition. Current findings may help clinicians differentiate word recognition errors due to language background from hearing pathologies.

Phonological competition within the word: Evidence from the phoneme similarity effect in spoken production

Theories of spoken production have not specifically addressed whether the phonemes of a word compete with each other for selection during phonological encoding (e.g., whether /t/ competes with /k/ in cat). Spoken production theories were evaluated and found to fall into three classes, theories positing (1) no competition, (2) competition among phonemes within the same syllable position, and (3) competition among all phonemes in a word. These predictions were tested by examining the effect of within-word phoneme similarity on oral reading reaction times using mixed-effects regression. Subjects took longer to begin uttering words containing similar phonemes than with dissimilar phonemes. This was true for consonant pairs in the onset, in the onset and coda, and in the onset and suffix. The results are most compatible with theories allowing all phonemes in a word to compete with each other. The possible relationship between these results and cross-linguistic patterns are also discussed.

Error Minimization in Phoneme based Automated Speech Recognition for Similar Sounding Phonemes

The technology of ASR (Automated Speech Recognition) has been quite successful with the use of Hidden Markov Model (HHM) with the aid of probabilistic and the best path methods. The words with limited vocabulary content can easily be modeled and trained. The modules with a large dictionary has to be modeled with context independent phones joining together to form a word. The major problem lies in recognizing the word with similar sounding phonemes. This paper aims in minimizing the error with similar sounding phonemes by using the viterbi algorithm for each similar syllable in the backward direction.