FAIC Foreign Accent Imitation Corpus Sara Neuhauser University of - - PowerPoint PPT Presentation

FAIC – Foreign Accent Imitation Corpus

FAIC – Foreign Accent Imitation Corpus

Sara Neuhauser

University of Jena, Germany

IAFPA 2011 Vienna, 24.–28.07.2011

FAIC – Foreign Accent Imitation Corpus Outline

1 Background

Preliminary study

2 Corpus

Speakers and accents Texts and recordings Data processing

3 Summary 4 References

FAIC – Foreign Accent Imitation Corpus Background

Background

imitation of foreign accents as a possible form of voice disguise alleged aims: conceal identity and mislead investigation

FAIC – Foreign Accent Imitation Corpus Background

Experimental studies have focused on

1 authenticity judgements by listeners

(e.g. Tate, 1979; Markham, 1999; Neuhauser&Simpson, 2007)

2 production of imitated foreign accents / different regional

dialects

(e.g. Torstensson et al. 2004; Neuhauser, 2008)

FAIC – Foreign Accent Imitation Corpus Background

FAIC [feIk] – Foreign Accent Imitation Corpus

http://www.uni-jena.de/˜x0nesa/faic.html

corpus collected within my PhD project (Neuhauser, 2011)

investigating foreign accent imitation from a production and a perception perspective within a German context

accent authenticity, accent identification VOT and voicing in plosives syllable initial glottalisation and glottal friction realisation of final /-@n/

published via an internet platform

examples of the recordings, all texts, description of data acquisition, recording conditions and data processing

FAIC – Foreign Accent Imitation Corpus Background Preliminary study

Preliminary study

determine which accents are claimed to be the easiest to imitate though it is of forensic interest which accents are most “popular”for accent imitation the forensic background of this study was deliberately withheld

exclude stereotypes

acquisition of German subjects (potential imitators) for the main investigation

FAIC – Foreign Accent Imitation Corpus Background Preliminary study

2 groups of native German speakers were asked to name the two accents they think would be the easiest to imitate

1 121 subj. (60 male, 61 female; mean age = 22.7 years) 2 70 subj. (27 male, 43 female; mean age = 21.9 years)

1st group: a list of accents were given

(AE, BE, French, Ital., Russ., Swiss Germ., Slav., Span., Turk., Vietn.)

choose up to 2 accents from this list or name another accent

control group: no accents were given

FAIC – Foreign Accent Imitation Corpus Background Preliminary study

French English Russian Turkish Italian Swiss Germ. Spanish Austrian Germ. Slavonic

20 40 60 80 100 120 140

Accents Number of votes

FAIC – Foreign Accent Imitation Corpus Corpus Speakers and accents

Corpus: Speakers and Accents

35 native German speakers imitating foreign accents Accent imitations No of speakers French accent 22 (15 female, 7 male) Am.-English accent 11 (5 female, 6 male) Russian accent 2 (1 female, 1 male) Turkish accent 1 (female) Total 351 (22 female, 13 male) mean age = 23.8 years mostly from eastern central Germany (Thuringia, Saxonia)

11 male speaker produced a French and an Am.-Engl. accent

FAIC – Foreign Accent Imitation Corpus Corpus Speakers and accents

12 non-native German speakers speaking German Mother tongue No of speakers French 4 (2 female, 2 male) Am.-English 8 (4 female, 4 male) Total 12 (6 female, 6 male)2 mean age = 23 years subjects differ in their L2 competence level noticeable foreign accent in German

2recordings of 1 Russian, 1 Czech, 1 Italian and 1 Am.-Engl. speaker

were excluded

FAIC – Foreign Accent Imitation Corpus Corpus Texts and recordings

Texts

2 prepared texts (about 220-240 words) ca. 1 week before the recordings composition of an own text with given words and phrases immediately before the recordings: 1 unprepared text spontaneous speech native Germans: disguised and undisguised productions (12 texts/speaker) non-native Germans: German productions and 1 text in their mother tongue (10 texts/speaker)

FAIC – Foreign Accent Imitation Corpus Corpus Texts and recordings

each subj. attended 2 recording sessions with a time-lag

• f 2 weeks

recording conditions and order of reading the texts were the same in both recording sessions speakers were informed about the forensic context immediately before the 1st recording session instruction to speak as naturally as possible rather than trying to produce standard German in the undisguised text versions

FAIC – Foreign Accent Imitation Corpus Corpus Texts and recordings

Recordings

sound treated room directly to PC digitizing at a sampling rate of 16 kHz and an amplitude resolution of 16 bits speaking distance to the microphone = ca. 40 cm

FAIC – Foreign Accent Imitation Corpus Corpus Texts and recordings

Examples

French accent:

female native German speaker reading text 1 female native French speaker reading text 1

American-English accent:

male native Am.-English speaker – spontaneous speech

FAIC – Foreign Accent Imitation Corpus Corpus Data processing

Data processing

French and Am.-Engl. accents

transliteration of the self-composed texts and the spontaneous speech annotation at the word level (506 of 518 signals)

using praat and xassp (converted to praat .TextGrid)

anonymisation of spontaneous speech

most significant personal details of each speaker (names, address, place of birth etc.) using a script and the annotation 400 Hz low pass filter, transition bandwidth 100 Hz – preserving pitch and temporal aspects

FAIC – Foreign Accent Imitation Corpus Corpus Data processing

Example

Annotation & Anonymisation

male native German speaker imitating Am.-Engl. accent (spontaneous speech)

FAIC – Foreign Accent Imitation Corpus Summary

Summary

FAIC offers annotated speech material of 47 native and non-native German speakers producing... imitated and authentic foreign accents read and spontaneous speech self-composed texts

FAIC – Foreign Accent Imitation Corpus Summary

FAIC is of forensic interest... detailed phonetic analyses of imitated and authentic accents produced in different linguistic activities (e.g. read vs. spontaneous) non-phonetic linguistic analyses of self-composed texts and spontaneous speech (e.g. syntactical or lexical aspects) analysis of speakers’ consistency during accent imitation (2 recording sessions)

FAIC – Foreign Accent Imitation Corpus Summary

Projects using FAIC

Neuhauser (2011) etc. Production and perception of imitated foreign accents Simpson & Neuhauser (2010) Epiphenomenal sound production in foreign accent disguise Witecy (2010) Foreign Accent Syndrome and the perception of accents Hellmann (2008) German as a foreign language for native American-English speakers

FAIC – Foreign Accent Imitation Corpus Summary

FAIC [feIk] – Foreign Accent Imitation Corpus http://www.uni-jena.de/˜x0nesa/faic.html

• r mail to

sara.neuhauser@uni-jena.de

Thanks to Adrian P. Simpson and to the student assistants Susanne Friedrich, Judith Hellmann, Anne-Katrin Leich and Toni Linke

FAIC – Foreign Accent Imitation Corpus References

Hellmann, J. (2008) Phonetische und phonologische Transferenzen und Interferenzen der Lernersprache amerikanischer Deutschlerner. (MA Thesis, University of Jena) Markham, D. (1999) Listeners and disguised voices: the imitation and perception of dialectal accent. IJSLL 6.2, 289–299 Neuhauser, S. (2011) Phonetische und linguistische Aspekte der Akzentimitation im forensischen Kontext. Produktion und Perzeption. (PhD Thesis, University of Jena) Neuhauser, S. (2008) Voice disguise using a foreign accent – phonetic and linguistic variation. IJSLL 15.2, 131–159 Neuhauser, S. & A.P. Simpson (2007) Imitated or authentic? Listeners judgements of foreign accents. Proc. 16th Saarbr¨ ucken, 1805–1808 Simpson, A.P. & S. Neuhauser (2010) The persistency of epiphenomenal sound production in foreign accent disguise. Proc. 19th IAFPA 2010, Trier, 31 Tate, D.A. (1979) Preliminary data on dialect in speech disguise. Current issues in the phonetic sciences: Proc. IPS-77 Congress (9). Amsterdam: Benjamins, 847–850 Torstensson, N., E.J. Eriksson, K.P.H.Sullivan (2004). Mimicked accents – Do speakers have similar cognitive prototypes? Proc. 10th Australian International Conference on Speech Science & Technology, Sydney, 271–276 Witecy, B. (2010) Akzentbeurteilung bei Foreign Accent Syndrome. (MA Thesis, University of Marburg)