39
An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida [email protected] [email protected]

An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida [email protected] [email protected]

Embed Size (px)

Citation preview

Page 1: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

An acoustic study of Dimasa tones

Priyankoo Sarmah and Caroline WiltshireUniversity of Florida

[email protected]@ufl.edu

Page 2: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

The Dimasa Language

Tibeto-Burman Language of the Bodo-Garo family.

Spoken by 88,543 people in Assam and Nagaland.(1991 Indian census report)

RCILTS, IIT Guwahati (http://www.iitg.ernet.in/rcilts/dimasa.htm )

2 tones: High, Unmarked Level

Singha (2003) 3 tones: High, Low and Unmarked Level

Page 3: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

This study…

How many tones are there in Dimasa?

What are the acoustic properties of Dimasa tones?

How do tones operate in monosyllables and disyllables?

Page 4: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Wordlist and speakers A list of segmentally homophonic words was constructed using data

from a native speaker of Dimasa.

The words were arranged in random order in the list.

10 native speakers of Dimasa (5 male, 5 female) were asked to produce the words in a sentence frame.

ang X thiba

Each word was repeated four times by the speakers.

Only the first three iterations were included in analysis to avoid listing effect.

Page 5: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Recording the data Recorder: Marantz PMD660 solid state

Sampling frequency: 48 Khz

Microphone: Audio-Technica AT4041

Storage: HP compact flash card

Page 6: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Analyzing the data Praat v.4.5.04

Measure 1: Track the pitch on the TBU at every 2% of the length of the TBU

Measure 2: The initiation point of the pitch on TBU

Measure 3: 20ms after initiation of the pitch on TBU

Page 7: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Analyzing the data Measure 1 is taken to track the pitch of the TBU

in detail. Automatic pitch tracking of Praat is not detailed

enough. It also normalizes the length as % values are derived.

Measure 2 and Measure 3 are taken to see if and to what extent the consonant in the onset affects the pitch of the TBU.

Page 8: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Analyzing the data Measure 1: Every 2% of the pitch on TBU = 50 points

TBU0

500

100

200

300

400

Time (s)0 0.272542

Page 9: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology: Analyzing the data Measure 2: Point of initiation of the pitch on the TBU (Pi) Measure 3: 20ms after the point of initiation (Pi+20)

20ms TBU 20ms

0

500

100

200

300

400

Time (s)0 0.272542

Page 10: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Methodology

For all the measurements specific scripts were written.

The script were run on 103 underived words of Dimasa.

The 103 words included monosyllabic and polysyllabic words.

Page 11: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Measure 2 and 3

Pi and Pi+20 are compared statistically

A standard t-test (p<0.05) is conducted on the data.

Grouped by the onset consonant.

Page 12: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 2 and 3

Vowels b d

0.070238755 5.53732E-14 3.06496E-08

dz g h

0.026574588 0.029406493 3.09369E-08

kh l m

4.03054E-06 2.20243E-05 0.089387967

r s sh

0.000723691 5.13417E-05 1.39625E-12

th w z

1.22493E-10 0.826479848 0.000801464

Page 13: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 2 and 3

[ph], [kh], [b], [d], [g], [dz], [h], [l], [r], [s], [sh], [z]

Showed significant effect on the following pitch.

[m] and [w] did not show significant effect on the following pitch.

Pi – (Pi+20) showed that [ph], [kh], [b], [d], [dz], [h], [l], [s], [sh], [z] raises the following

pitch.

[r] depresses the following pitch.

Page 14: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

Pitch points were calculated on 50 points across the pitch of each TBU (every 2%)

These 50 points were replotted on a graph to reconstruct the pitch track.

Page 15: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

Initial examination of data set in Singha (2003):

thi ‘speak’ thi ‘to die’ thi ‘blood’

lai ‘page’ lai ‘easy’ lai ‘wish’

maithai‘year’ maithai‘crop’ maithai ‘source’

Page 16: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1 Demonstrates tonal distinction in segmentally

homogeneous pairs./thi/

150

170

190

210

230

250

270

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Point no.

Freq

uenc

y (H

z) SPEAK

BLOOD

DIE

Page 17: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1/lai/

150

160

170

180

190

200

210

220

230

240

250

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Point no.

Freq

uenc

y (H

z) EASY

WISH

PAGE

Page 18: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

/maithai/ ‘crop’

a ng m ai th ai th ib a

0

500

100

200

300

400

Time (s)0 0.918167

Page 19: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

/maithai/ ‘source’

a ng m ai th ai th i b a

0

500

100

200

300

400

Time (s)0 0.951438

Page 20: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

/maithai/ ‘year’

a ng m ai th ai th ib a

0

500

100

200

300

400

Time (s)0 1.15398

Page 21: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

Compared to Singha (2003):

/thi/ shows evidence for three distinct tones

/lai/ does not show evidence for three tones

/maithai/ does not show evidence for three tones.

Page 22: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

Another interesting set:

bai ‘to break’ bai ‘to cross’ bai ‘to dance’ bai ‘to order’ bai ‘to ship’ bai ‘to spin’

Page 23: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

/bai/

150

170

190

210

230

250

270

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Point no.

Freq

uenc

y (H

z)

CROSS

ORDER

DANCE

FILTER

BREAK

SPIN

Page 24: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

In most of the cases:/wai/

150

170

190

210

230

250

270

290

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Point no.

Freq

uenc

y (H

z)

FIRE

CHEW

Page 25: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Measure 1

Tones: Dimasa has three lexical tones:

A high tone (rising) A level mid tone A low tone (falling)

Page 26: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Bisyllables in Dimasa

No mention in Singha (2003)

Bodo-Garo languages primarily show tone assignment on only one syllable.

(Joseph and Burling, 2001; Sarmah 2003)

Page 27: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Bisyllables in Dimasa

goron company goron to confuse

miya bamboo shoot miya male person miya yesterday

Page 28: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Bisyllables

/goron/ ‘company’

a ng h a d i th ib a0

500

100

200

300

400

Time (s)0 1.11777

a ng h a d i th ib a0

500

100

200

300

400

Time (s)0 1.13108

ang h a z i th ib a0

500

100

200

300

400

Time (s)0 1.11525

ang h a z i th ib a0

500

100

200

300

400

Time (s)0 1.14958

a ng g o r o n th ib a0

500

100

200

300

400

Time (s)0 1.54158

a ng g o r o n th i b a0

500

100

200

300

400

Time (s)0 1.25083

Page 29: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Bisyllables

/goron/ ‘confuse’

a ng h a d i th ib a0

500

100

200

300

400

Time (s)0 1.11777

a ng h a d i th ib a0

500

100

200

300

400

Time (s)0 1.13108

ang h a z i th ib a0

500

100

200

300

400

Time (s)0 1.11525

ang h a z i th ib a0

500

100

200

300

400

Time (s)0 1.14958

a ng g o r o n th ib a0

500

100

200

300

400

Time (s)0 1.54158

a ng g o r o n th i b a0

500

100

200

300

400

Time (s)0 1.25083

Page 30: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

miya ‘male person’

a ng m i j a th i b a

0

500

100

200

300

400

Time (s)0 1.11527

Page 31: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

miya ‘bamboo shoot’

a ng m i j a th i b a

0

500

100

200

300

400

Time (s)0 1.09115

Page 32: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

miya ‘yesterday’

a ng m i j a h th i b a

0

500

100

200

300

400

Time (s)0 1.05615

Page 33: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Results: Bisyllables

Tone assigned only on the second syllable.

Similar phenomenon to Bodo (Sarmah 2003)

The first syllable is assigned a mid level tone.

Similar phenomenon in other 9 sets of bisyllables we examined.

Page 34: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Summary of results

Significant consonantal effects on pitch

3 tones in Dimasa A high tone (rising) A level mid tone A low tone (falling)

A mono/disyllabic word can have one and only one tone assigned to it.

Page 35: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Association of glottal stops with high tones.

Trisyllabic words.

A perception test

More data…

Page 36: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

References

Joseph, U. V., and Burling, Robbins. 2001. “Tone correspondences among the Bodo Languages. Linguistics of the Tibeto-Burman Area. 24.2: 41-55.

Sarmah, Priyankoo. 2003. Some Aspects of the Tonal Phonology of Bodo. M.Phil. dissertation, CIEFL, Hyderabad, India.

Singha, Dhiren. 2003. The Phonology & Morphology of Dimasa. M.A. dissertation, Assam University, Silchar, Assam

Page 37: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Acknowledgment

Mr. Uttam Bathari, Asst. Director, ICHR, Guwahati, Assam.

University of Florida, Graduate Research Program.

and…

Page 38: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

Acknowledgment

The Dimasa Students’ Association

Guwahati

Page 39: An acoustic study of Dimasa tones Priyankoo Sarmah and Caroline Wiltshire University of Florida priyanku@ufl.edu wiltshir@ufl.edu

This presentation is available: By email: [email protected]

Online: http://plaza.ufl.edu/priyanku/neils.ppt