Calbert Graham
2021
Graham, Calbert
L1 Identification from L2 Speech Using Neural Spectrogram Analysis Conference
2021.
@conference{Graham2021,
title = {L1 Identification from L2 Speech Using Neural Spectrogram Analysis},
author = {Calbert Graham},
doi = {10.21437/Interspeech.2021-1545},
year = {2021},
date = {2021-08-30},
urldate = {2021-08-30},
pages = {3959 - 3963},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Graham, Calbert
L1 Identification from L2 Speech Using Neural Spectrogram Analysis Journal Article
In: 2021.
@article{Graham2021b,
title = {L1 Identification from L2 Speech Using Neural Spectrogram Analysis},
author = {Calbert Graham},
year = {2021},
date = {2021-06-19},
urldate = {2021-06-19},
abstract = {It is well-known that the characteristics of L2 speech are highly influenced by the speakers’ L1. The main objective of this study was to uncover discriminative speech features to identify the L1 background of a speaker from their L2 English speech. Traditional phonetic approaches tend to compare speakers based on a pre-selected set of acoustic features, which may not be sufficient to capture all the unique traces of the L1 in the L2 speech for forensic speaker profiling purposes. Convolutional Neural Network (CNN) has the potential to remedy this issue through the automatic processing of the visual spectrogram. This paper reports a series of CNN classification experiments modelled on spectrogram images. The classification problem consisted of determining whether English speech samples are spoken by a native speaker of English, Japanese, Dutch, French, or Polish. Both phonetically transcribed and untranscribed speech data were used. Overall, results showed that the CNN achieved a high level of accuracy in identifying the speakers’ L1s based on spectrogram pictures without explicit phonetic segmentation. However, the results also showed that training the classifiers on certain combinations of phonetically modelled spectrogram images, which would make features more transparent, can produce results with comparable accuracy rates.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
It is well-known that the characteristics of L2 speech are highly influenced by the speakers’ L1. The main objective of this study was to uncover discriminative speech features to identify the L1 background of a speaker from their L2 English speech. Traditional phonetic approaches tend to compare speakers based on a pre-selected set of acoustic features, which may not be sufficient to capture all the unique traces of the L1 in the L2 speech for forensic speaker profiling purposes. Convolutional Neural Network (CNN) has the potential to remedy this issue through the automatic processing of the visual spectrogram. This paper reports a series of CNN classification experiments modelled on spectrogram images. The classification problem consisted of determining whether English speech samples are spoken by a native speaker of English, Japanese, Dutch, French, or Polish. Both phonetically transcribed and untranscribed speech data were used. Overall, results showed that the CNN achieved a high level of accuracy in identifying the speakers’ L1s based on spectrogram pictures without explicit phonetic segmentation. However, the results also showed that training the classifiers on certain combinations of phonetically modelled spectrogram images, which would make features more transparent, can produce results with comparable accuracy rates.
2020
Graham, Calbert; Laméris, Tim
L2 Perception and Production of Japanese Lexical Pitch: A Suprasegmental Similarity Account Journal Article
In: Journal of Monolingual and Bilingual Speech, vol. 2, no. 1, pp. 106-136, 2020.
@article{Graham2020,
title = {L2 Perception and Production of Japanese Lexical Pitch: A Suprasegmental Similarity Account},
author = {Calbert Graham and Tim Laméris},
doi = {10.1558/jmbs.14948},
year = {2020},
date = {2020-09-28},
urldate = {2020-09-28},
journal = {Journal of Monolingual and Bilingual Speech},
volume = {2},
number = {1},
pages = {106-136},
abstract = {Adults are known to have difficulties acquiring suprasegmental speech that involves pitch (f0) in a second language (L2) (Graham & Post, 2018; Hirata, 2015; Wang, Spence, Jongman & Sereno, 1999; Wong & Perrachione, 2007). Previous research has suggested that the perceived similarity between L1 and L2 phonology may influence how easily segmental speech is acquired, but this notion of ‘similarity’ may also apply to suprasegmental speech (So & Best, 2010; Wu, Munro & Wang, 2014). In this paper, the L2 acquisition of Japanese lexical pitch was assessed under a ‘Suprasegmental Similarity Account’, which is a theoretical framework inspired by previous models of segmental and suprasegmental speech (Best & Tyler, 2007; Flege, 1995; Mennen, 2015) to account for the L2 acquisition of word prosody. Eight adult native speak- ers of Japanese and eight adult English-native advanced learners of Japanese participated in a perception and production study of Japanese lexical pitch patterns. Both groups performed similarly in perception, but non-native speakers performed significantly worse in production, particularly for ‘unac- cented’ Low–High–High patterns. These findings are discussed in light of the ‘Suprasegmental Similarity Account’. Keywords: Japanese, lexical pitch, L2 speech learning, perception, production},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Adults are known to have difficulties acquiring suprasegmental speech that involves pitch (f0) in a second language (L2) (Graham & Post, 2018; Hirata, 2015; Wang, Spence, Jongman & Sereno, 1999; Wong & Perrachione, 2007). Previous research has suggested that the perceived similarity between L1 and L2 phonology may influence how easily segmental speech is acquired, but this notion of ‘similarity’ may also apply to suprasegmental speech (So & Best, 2010; Wu, Munro & Wang, 2014). In this paper, the L2 acquisition of Japanese lexical pitch was assessed under a ‘Suprasegmental Similarity Account’, which is a theoretical framework inspired by previous models of segmental and suprasegmental speech (Best & Tyler, 2007; Flege, 1995; Mennen, 2015) to account for the L2 acquisition of word prosody. Eight adult native speak- ers of Japanese and eight adult English-native advanced learners of Japanese participated in a perception and production study of Japanese lexical pitch patterns. Both groups performed similarly in perception, but non-native speakers performed significantly worse in production, particularly for ‘unac- cented’ Low–High–High patterns. These findings are discussed in light of the ‘Suprasegmental Similarity Account’. Keywords: Japanese, lexical pitch, L2 speech learning, perception, production
Graham, Calbert
Late bilingual acquisition of narrow corrective focus in American English by Japanese and Spanish native speakers Journal Article
In: The Journal of the Acoustical Society of America, 2020.
@article{Graham2020b,
title = {Late bilingual acquisition of narrow corrective focus in American English by Japanese and Spanish native speakers},
author = {Calbert Graham},
year = {2020},
date = {2020-09-06},
urldate = {2020-09-06},
journal = {The Journal of the Acoustical Society of America},
abstract = {The present study examined the non-native acquisition of English tonal patterns and the phonetic implementation (i.e. duration and pitch range realisations) in corrective narrow focus marking by native speakers of Spanish (a typologically close language to English) and Japanese (a typologically more distant language to English). The study explores to what extent transfer of native language properties can explain L2 intonation patterns in narrow focus marking, and whether intonation is learned in a uniform fashion in terms of its form, function, and rate of developmental progression. Findings indicate that the similarity of the source and target languages (both in terms of typological closeness in their intonation systems and the actual phonetic manifestation of narrow focus) may influence L2 learners' realisation of narrow focus intonation. However, in terms of the phonetic implementation of cues to signal narrow focus the f0 range cue appeared to pose the most difficulty with learners producing a narrower pitch range than the native speakers, regardless of their English proficiency or L1. The duration cue was found to be problematic only for the Japanese learners of English. Overall, the study reveals that the acquisition of the phonological patterns of narrow focus intonation depends on the similarity between the L1 and the L2 in both surface tonal structure and phonological rules that constrain it. We argue that L2 learners have to acquire both aspects, as they learn to master the various factors that determine what the appropriate language-specific phonological structures and phonetic implementation patterns are in the L2.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The present study examined the non-native acquisition of English tonal patterns and the phonetic implementation (i.e. duration and pitch range realisations) in corrective narrow focus marking by native speakers of Spanish (a typologically close language to English) and Japanese (a typologically more distant language to English). The study explores to what extent transfer of native language properties can explain L2 intonation patterns in narrow focus marking, and whether intonation is learned in a uniform fashion in terms of its form, function, and rate of developmental progression. Findings indicate that the similarity of the source and target languages (both in terms of typological closeness in their intonation systems and the actual phonetic manifestation of narrow focus) may influence L2 learners' realisation of narrow focus intonation. However, in terms of the phonetic implementation of cues to signal narrow focus the f0 range cue appeared to pose the most difficulty with learners producing a narrower pitch range than the native speakers, regardless of their English proficiency or L1. The duration cue was found to be problematic only for the Japanese learners of English. Overall, the study reveals that the acquisition of the phonological patterns of narrow focus intonation depends on the similarity between the L1 and the L2 in both surface tonal structure and phonological rules that constrain it. We argue that L2 learners have to acquire both aspects, as they learn to master the various factors that determine what the appropriate language-specific phonological structures and phonetic implementation patterns are in the L2.
2019
Graham, Calbert; Nolan, Francis
Articulation Rate as a Metric in Spoken Language Assessment Conference
2019.
@conference{Graham2019,
title = {Articulation Rate as a Metric in Spoken Language Assessment},
author = {Calbert Graham and Francis Nolan},
doi = {10.21437/Interspeech.2019-2098},
year = {2019},
date = {2019-07-15},
urldate = {2019-07-15},
abstract = {Automated evaluation of non-native pronunciation provides a consistent and more cost-efficient alternative to human evaluation. To that end, there is considerable interest in deriving metrics that are based on the cues human listeners use to judge pronunciation. Previous research reported the use of phonetic features such as vowel characteristics in automated spoken language evaluation. The present study extends this line of work on the significance of phonetic features in automated evaluation of L2 speech (both assessment and feedback). Predictive modelling techniques examined the relationship between various articulation rate metrics one the one hand, and the proficiency and L1 background of non-native English speakers on the other. It was found that the optimal predictive model was one in which the phonetic details of phoneme articulation were factored in the analysis of articulation rate. Model performance varied also according to the L1 background of speakers. The implications for assessment and feedback are discussed.
},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Automated evaluation of non-native pronunciation provides a consistent and more cost-efficient alternative to human evaluation. To that end, there is considerable interest in deriving metrics that are based on the cues human listeners use to judge pronunciation. Previous research reported the use of phonetic features such as vowel characteristics in automated spoken language evaluation. The present study extends this line of work on the significance of phonetic features in automated evaluation of L2 speech (both assessment and feedback). Predictive modelling techniques examined the relationship between various articulation rate metrics one the one hand, and the proficiency and L1 background of non-native English speakers on the other. It was found that the optimal predictive model was one in which the phonetic details of phoneme articulation were factored in the analysis of articulation rate. Model performance varied also according to the L1 background of speakers. The implications for assessment and feedback are discussed.
Graham, Calbert; Nolan, Francis
Articulation rate as a metric in spoken language assessment Bachelor Thesis
2019.
@bachelorthesis{Graham2019b,
title = {Articulation rate as a metric in spoken language assessment},
author = {Calbert Graham and Francis Nolan},
year = {2019},
date = {2019-07-15},
urldate = {2019-07-15},
keywords = {},
pubstate = {published},
tppubtype = {bachelorthesis}
}
Graham, Calbert; Post, Brechtje
Constancy and Variation in Speech: Phonetic Realisation and Abstraction Journal Article
In: Phonetica, vol. 76, no. 2-3, pp. 87-99, 2019.
@article{Graham2019c,
title = {Constancy and Variation in Speech: Phonetic Realisation and Abstraction},
author = {Calbert Graham and Brechtje Post},
doi = {10.1159/000497439},
year = {2019},
date = {2019-05-01},
urldate = {2019-05-01},
journal = {Phonetica},
volume = {76},
number = {2-3},
pages = {87-99},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2018
Graham, Calbert; rechtje Post,
Second language acquisition of intonation: Peak alignment in American English Journal Article
In: Journal of Phonetics, vol. 66, pp. 1-14, 2018.
@article{Graham2018,
title = {Second language acquisition of intonation: Peak alignment in American English},
author = {Calbert Graham and rechtje Post},
doi = {10.1016/j.wocn.2017.08.002},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
journal = {Journal of Phonetics},
volume = {66},
pages = {1-14},
abstract = {The objective of the present study was to investigate (1) whether, and to what degree, late bilinguals of different L1 backgrounds are comparable to native speakers in the phonetic implementation of tonal targets in their L2, (2) whether they exhibit general patterns of acquisition irrespective of the typological closeness of their L1 to their L2, and (3) whether learners’ choice of accent contours and the alignment of the high tone (H∗) proceeds in parallel with proficiency in the L2. More specifically, we examined the acquisition of the nuclear contour composition and the H∗ alignment of the American English (L)H∗L- (i.e. pitch accent and boundary tone combination) in initial-stressed and final-stressed words by Japanese and Spanish late bilingual speakers at varying proficiency levels in American English. Our results show that the L1 Spanish speakers were more comparable than the L1 Japanese speakers to the native English speakers in the phonological aspect of intonation (choice of pitch accent contour). In terms of peak alignment, we found that the late bilinguals generally tended to realise significantly later alignment than the native speakers, although the precise manifestation of this varied according to the L1 background of speakers and the stress pattern of words.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The objective of the present study was to investigate (1) whether, and to what degree, late bilinguals of different L1 backgrounds are comparable to native speakers in the phonetic implementation of tonal targets in their L2, (2) whether they exhibit general patterns of acquisition irrespective of the typological closeness of their L1 to their L2, and (3) whether learners’ choice of accent contours and the alignment of the high tone (H∗) proceeds in parallel with proficiency in the L2. More specifically, we examined the acquisition of the nuclear contour composition and the H∗ alignment of the American English (L)H∗L- (i.e. pitch accent and boundary tone combination) in initial-stressed and final-stressed words by Japanese and Spanish late bilingual speakers at varying proficiency levels in American English. Our results show that the L1 Spanish speakers were more comparable than the L1 Japanese speakers to the native English speakers in the phonological aspect of intonation (choice of pitch accent contour). In terms of peak alignment, we found that the late bilinguals generally tended to realise significantly later alignment than the native speakers, although the precise manifestation of this varied according to the L1 background of speakers and the stress pattern of words.
2016
Graham, Calbert
Implicit learning of Latin stress regularities Journal Article
In: Studies in Second Language Acquisition, vol. 40, no. 1, 2016.
@article{Graham2016b,
title = {Implicit learning of Latin stress regularities},
author = {Calbert Graham},
doi = {10.1017/S0272263116000371},
year = {2016},
date = {2016-12-01},
urldate = {2016-12-01},
journal = {Studies in Second Language Acquisition},
volume = {40},
number = {1},
abstract = {This study examines whether Japanese subjects can implicitly learn stress pattern regularities, not present in their L1, after a brief auditory exposure. In the exposure phase, the subjects listened to words bearing stress patterned after Latin, but with a highly restricted consonant inventory. They performed a judgement task designed to test whether they were able to extract stress pattern regularities from the brief listening experience. We assessed participants’ awareness of the inherent stress regularities by analysing confidence in making decisions and by verbal report. Results suggest that although participants remained unaware of the underlying stress regularities they performed significantly above chance in endorsing correctly stressed novel items even when they claimed to be guessing. In addition, there was no difference in confidence between correct and incorrect judgments. These results suggest that brief exposure resulted in implicit knowledge of abstract stress assignment rules. However, participants rejected correctly stressed words that contained consonants that were not present in the test phase. On the basis of this finding, we speculate that in the acquisition of second language phonology, a violation at the segmental level may be deemed more costly than a violation in the prosodic domain.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study examines whether Japanese subjects can implicitly learn stress pattern regularities, not present in their L1, after a brief auditory exposure. In the exposure phase, the subjects listened to words bearing stress patterned after Latin, but with a highly restricted consonant inventory. They performed a judgement task designed to test whether they were able to extract stress pattern regularities from the brief listening experience. We assessed participants’ awareness of the inherent stress regularities by analysing confidence in making decisions and by verbal report. Results suggest that although participants remained unaware of the underlying stress regularities they performed significantly above chance in endorsing correctly stressed novel items even when they claimed to be guessing. In addition, there was no difference in confidence between correct and incorrect judgments. These results suggest that brief exposure resulted in implicit knowledge of abstract stress assignment rules. However, participants rejected correctly stressed words that contained consonants that were not present in the test phase. On the basis of this finding, we speculate that in the acquisition of second language phonology, a violation at the segmental level may be deemed more costly than a violation in the prosodic domain.
Graham, Calbert
Vowel characteristics in the assessment of L2 intelligibility Conference
2016.
@conference{Graham2016,
title = {Vowel characteristics in the assessment of L2 intelligibility},
author = {Calbert Graham},
year = {2016},
date = {2016-09-09},
urldate = {2016-09-09},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Graham, Calbert; Buttery, Paula; Nolan, Francis
Vowel Characteristics in the Assessment of L2 English Pronunciation Conference
2016.
@conference{Graham2016c,
title = {Vowel Characteristics in the Assessment of L2 English Pronunciation},
author = {Calbert Graham and Paula Buttery and Francis Nolan},
doi = {10.21437/Interspeech.2016-1630},
year = {2016},
date = {2016-09-08},
urldate = {2016-09-08},
pages = {1127 - 1131},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Graham, Calbert; Caines, Andrew; Bentz, Christian; Buttery, Paula; Polzehl, Tim
Crowdsourcing a multilingual speech corpus: recording, transcription and annotation of the CROWDED CORPUS Conference
2016.
@conference{Graham2016d,
title = {Crowdsourcing a multilingual speech corpus: recording, transcription and annotation of the CROWDED CORPUS},
author = {Calbert Graham and Andrew Caines and Christian Bentz and Paula Buttery and Tim Polzehl},
year = {2016},
date = {2016-03-15},
urldate = {2016-03-15},
abstract = {We announce the release of the CROWDED CORPUS: a pair of speech corpora collected via crowdsourcing, containing a native speaker corpus of English (CROWDED_ENGLISH), and a corpus of German/English bilinguals (CROWDED_BILINGUAL). Release 1 of the CROWDED CORPUS contains 1000 recordings amounting to 33,400 tokens collected from 80 speakers and is freely available to other researchers. We recruited participants via the Crowdee application for Android. Recruits were prompted to respond to business-topic questions of the type found in language learning oral tests. We then used the CrowdFlower web application to pass these recordings to crowdworkers for transcription and annotation of errors and sentence boundaries. Finally, the sentences were tagged and parsed using standard natural language processing tools. We propose that crowdsourcing is a valid and economical method for corpus collection, and discuss the advantages and disadvantages of this approach.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
We announce the release of the CROWDED CORPUS: a pair of speech corpora collected via crowdsourcing, containing a native speaker corpus of English (CROWDED_ENGLISH), and a corpus of German/English bilinguals (CROWDED_BILINGUAL). Release 1 of the CROWDED CORPUS contains 1000 recordings amounting to 33,400 tokens collected from 80 speakers and is freely available to other researchers. We recruited participants via the Crowdee application for Android. Recruits were prompted to respond to business-topic questions of the type found in language learning oral tests. We then used the CrowdFlower web application to pass these recordings to crowdworkers for transcription and annotation of errors and sentence boundaries. Finally, the sentences were tagged and parsed using standard natural language processing tools. We propose that crowdsourcing is a valid and economical method for corpus collection, and discuss the advantages and disadvantages of this approach.
2015
Graham, Calbert; Moore, Russell; Caines, Andrew; Buttery, Paula
Incremental Dependency Parsing and Disfluency Detection in Spoken Learner English Conference
vol. 9302, 2015, ISBN: 978-3-319-24032-9.
@conference{Graham2015b,
title = {Incremental Dependency Parsing and Disfluency Detection in Spoken Learner English},
author = {Calbert Graham and Russell Moore and Andrew Caines and Paula Buttery},
doi = {10.1007/978-3-319-24033-6_53},
isbn = {978-3-319-24032-9},
year = {2015},
date = {2015-09-14},
urldate = {2015-09-14},
volume = {9302},
pages = {470 - 479},
abstract = {This paper investigates the suitability of state-of-the-art natural language processing (NLP) tools for parsing the spoken language of second language learners of English. The task of parsing spoken learner-language is important to the domains of automated language assessment (ALA) and computer-assisted language learning (CALL). Due to the non-canonical nature of spoken language (containing filled pauses, non-standard grammatical variations, hesitations and other disfluencies) and compounded by a lack of available training data, spoken language parsing has been a challenge for standard NLP tools. Recently the Redshift parser (Honnibal et al. In: Proceedings of CoNLL (2013)) has been shown to be successful in identifying grammatical relations and certain disfluencies in native speaker spoken language, returning unlabelled dependency accuracy of 90.5% and a disfluency F-measure of 84.1% (Honnibal & Johnson: TACL 2, 131-142 (2014)). We investigate how this parser handles spoken data from learners of English at various proficiency levels. Firstly, we find that Redshift’s parsing accuracy on non-native speech data is comparable to Honnibal & Johnson’s results, with 91.1% of dependency relations correctly identified. However, disfluency detection is markedly down, with an F-measure of just 47.8%. We attempt to explain why this should be, and investigate the effect of proficiency level on parsing accuracy. We relate our findings to the use of NLP technology for CALL and ALA applications.
},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
This paper investigates the suitability of state-of-the-art natural language processing (NLP) tools for parsing the spoken language of second language learners of English. The task of parsing spoken learner-language is important to the domains of automated language assessment (ALA) and computer-assisted language learning (CALL). Due to the non-canonical nature of spoken language (containing filled pauses, non-standard grammatical variations, hesitations and other disfluencies) and compounded by a lack of available training data, spoken language parsing has been a challenge for standard NLP tools. Recently the Redshift parser (Honnibal et al. In: Proceedings of CoNLL (2013)) has been shown to be successful in identifying grammatical relations and certain disfluencies in native speaker spoken language, returning unlabelled dependency accuracy of 90.5% and a disfluency F-measure of 84.1% (Honnibal & Johnson: TACL 2, 131-142 (2014)). We investigate how this parser handles spoken data from learners of English at various proficiency levels. Firstly, we find that Redshift’s parsing accuracy on non-native speech data is comparable to Honnibal & Johnson’s results, with 91.1% of dependency relations correctly identified. However, disfluency detection is markedly down, with an F-measure of just 47.8%. We attempt to explain why this should be, and investigate the effect of proficiency level on parsing accuracy. We relate our findings to the use of NLP technology for CALL and ALA applications.
Graham, Calbert
Phonetic and prosodic features in automated spoken language assessment Conference
2015.
@conference{Graham2015c,
title = {Phonetic and prosodic features in automated spoken language assessment},
author = {Calbert Graham},
year = {2015},
date = {2015-08-12},
urldate = {2015-08-12},
abstract = {The traditional approach of using trained human assessors to evaluate spoken language proficiency has proven to be both expensive and time-consuming. This has led to renewed effort to develop tools for the automated assessment of non-native speech. Previous work on automated assessment has generally not focused on acoustic features, although it has long been established that listeners use such features in judging the naturalness of L2 pronunciation. Including such features in assessment will go a long way in making the implicit knowledge examiners use in assessing pronunciation more explicit, and can be very useful, for instance, in a computer-assisted language learning (CALL) pronunciation training programme. In this paper, we describe one strand of research in our ALTA project and discuss the role of phonetic features in automated assessment of non-native speech. We will briefly discuss some of the challenges we face in automatically measuring phonetic/prosodic features, and how we go about striking a compromise between what may be a linguistically meaningful feature and what is actually measurable given the constraints of our system.
},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
The traditional approach of using trained human assessors to evaluate spoken language proficiency has proven to be both expensive and time-consuming. This has led to renewed effort to develop tools for the automated assessment of non-native speech. Previous work on automated assessment has generally not focused on acoustic features, although it has long been established that listeners use such features in judging the naturalness of L2 pronunciation. Including such features in assessment will go a long way in making the implicit knowledge examiners use in assessing pronunciation more explicit, and can be very useful, for instance, in a computer-assisted language learning (CALL) pronunciation training programme. In this paper, we describe one strand of research in our ALTA project and discuss the role of phonetic features in automated assessment of non-native speech. We will briefly discuss some of the challenges we face in automatically measuring phonetic/prosodic features, and how we go about striking a compromise between what may be a linguistically meaningful feature and what is actually measurable given the constraints of our system.
Graham, Calbert
Fundamental Frequency Range in Japanese and English: The Case of Simultaneous Bilinguals Journal Article
In: Phonetica, vol. 71, no. 4, pp. 271-295, 2015.
@article{Graham2015,
title = {Fundamental Frequency Range in Japanese and English: The Case of Simultaneous Bilinguals},
author = {Calbert Graham},
doi = {10.1159/000381627},
year = {2015},
date = {2015-05-20},
journal = {Phonetica},
volume = {71},
number = {4},
pages = {271-295},
abstract = {This article reports an experiment to test whether Japanese and English speakers vary their fundamental frequency (f0) range as a function of the language spoken. Speech samples of Japanese-(American) English simultaneous bilinguals (5 males, 5 females; all undergraduates at UC Berkeley) performing comparable reading tasks in their two native languages were analysed. The study builds on a relatively new approach to measuring f0 range that computes its high and low points from tonal targets in the intonational phonology. Unlike in most previous studies where f0 range is traditionally treated as a one-dimensional measure, f0 range in this study is measured along two quasi-independent dimensions: level and span. Findings revealed statistically significant crosslanguage differences between Japanese and English in both dimensions. This provides new insights into the relation between prosodic structure and f0 range in these two typologically different prosodic systems. © 2015 S. Karger AG, Basel.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This article reports an experiment to test whether Japanese and English speakers vary their fundamental frequency (f0) range as a function of the language spoken. Speech samples of Japanese-(American) English simultaneous bilinguals (5 males, 5 females; all undergraduates at UC Berkeley) performing comparable reading tasks in their two native languages were analysed. The study builds on a relatively new approach to measuring f0 range that computes its high and low points from tonal targets in the intonational phonology. Unlike in most previous studies where f0 range is traditionally treated as a one-dimensional measure, f0 range in this study is measured along two quasi-independent dimensions: level and span. Findings revealed statistically significant crosslanguage differences between Japanese and English in both dimensions. This provides new insights into the relation between prosodic structure and f0 range in these two typologically different prosodic systems. © 2015 S. Karger AG, Basel.
Graham, Calbert; Nolan, Francis; Caines, Andrew; Buttery, Paula
Automated assessment of non-native speech using vowel formant features Conference
2015.
@conference{Graham2015d,
title = {Automated assessment of non-native speech using vowel formant features},
author = {Calbert Graham and Francis Nolan and Andrew Caines and Paula Buttery},
year = {2015},
date = {2015-01-01},
urldate = {2015-01-01},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
2013
Graham, Calbert; Post, Brechtje
Realisation of tonal alignment in the English of Japanese-English late bilinguals Conference
2013.
@conference{Graham2013,
title = {Realisation of tonal alignment in the English of Japanese-English late bilinguals},
author = {Calbert Graham and Brechtje Post},
doi = {10.21437/Interspeech.2013-557},
year = {2013},
date = {2013-08-25},
urldate = {2013-08-25},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Graham, Calbert; Post, Brechtje
Realisation of tonal alignment in the English of japanese-english late bilinguals Journal Article
In: Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, pp. 2390 - 2394, 2013.
@article{Graham2013b,
title = {Realisation of tonal alignment in the English of japanese-english late bilinguals},
author = {Calbert Graham and Brechtje Post},
year = {2013},
date = {2013-01-01},
urldate = {2013-01-01},
journal = {Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH},
pages = {2390 - 2394},
abstract = {Several factors have been attested to affect the temporal synchronisation of tonal targets such as syllable duration, segmental structure and proximity to word or intonational boundaries, e.g. [1], [2], [3]. Given the apparent language-specific nature of tonal alignment [4], it can be expected that late bilinguals who are acquiring a second language will need to learn the alignment implementation rules of that language, in addition to other aspects. This study compared the tonal alignment patterns of Japanese late bilingual English speakers and monolingual English speakers in order to investigate to what extent learners transfer their native implementation strategies to the interlanguage, and whether alignment changes with proficiency. The results show that, although initial-accented words were aligned later than final-accented words for all groups, as expected, the Japanese bilinguals aligned the former significantly later than the monolinguals. Further analyses revealed that their off-target realisations were generally limited to a specific type of syllable structure that we speculate may be linked to peak delay in their L1.These results are taken as evidence of prosodic transfer and suggest that late bilinguals will need to learn the L2 phonetic implementation rules of alignment independently of their acquisition of the phonology.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Several factors have been attested to affect the temporal synchronisation of tonal targets such as syllable duration, segmental structure and proximity to word or intonational boundaries, e.g. [1], [2], [3]. Given the apparent language-specific nature of tonal alignment [4], it can be expected that late bilinguals who are acquiring a second language will need to learn the alignment implementation rules of that language, in addition to other aspects. This study compared the tonal alignment patterns of Japanese late bilingual English speakers and monolingual English speakers in order to investigate to what extent learners transfer their native implementation strategies to the interlanguage, and whether alignment changes with proficiency. The results show that, although initial-accented words were aligned later than final-accented words for all groups, as expected, the Japanese bilinguals aligned the former significantly later than the monolinguals. Further analyses revealed that their off-target realisations were generally limited to a specific type of syllable structure that we speculate may be linked to peak delay in their L1.These results are taken as evidence of prosodic transfer and suggest that late bilinguals will need to learn the L2 phonetic implementation rules of alignment independently of their acquisition of the phonology.
Graham, Calbert
Revisiting f0 Range Production in Japanese-English Simultaneous Bilinguals Journal Article
In: Berkeley Phonology Lab Report, vol. 9, 2013.
@article{Graham2013c,
title = {Revisiting f0 Range Production in Japanese-English Simultaneous Bilinguals},
author = {Calbert Graham},
doi = {10.5070/P78WP244GC},
year = {2013},
date = {2013-01-01},
urldate = {2013-01-01},
journal = {Berkeley Phonology Lab Report},
volume = {9},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2011
Graham, Calbert; Galaczi, Evelina D.; Post, Brechtje; Li, Aike
Measuring L2 English Phonological Proficiency: Implications for Language Assessment Conference
2011.
@conference{Graham2011,
title = {Measuring L2 English Phonological Proficiency: Implications for Language Assessment},
author = {Calbert Graham and Evelina D. Galaczi and Brechtje Post and Aike Li},
year = {2011},
date = {2011-09-01},
urldate = {2011-09-01},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Graham, Calbert
A case study of Japanese acquisition of American English stress Journal Article
In: Journal of the Phonetic Society of Japan, vol. 15, 2011.
@article{Graham2011b,
title = {A case study of Japanese acquisition of American English stress},
author = {Calbert Graham},
year = {2011},
date = {2011-01-01},
urldate = {2011-01-01},
journal = {Journal of the Phonetic Society of Japan},
volume = {15},
keywords = {},
pubstate = {published},
tppubtype = {article}
}