汉语语句声调的降阶课件.ppt

1、 王茂林 Mao-Lin WANG Hua Wu Introduction In tone languages,downstep typically consists of the stepwise lowering of High(H)tones in certain contexts.In automatic downstep,H tones are lowered in sequences of alternating H and Low(L)tones.In nonautomatic downstep,often noted H!H,there is no overt conditio

2、ning L tone between the two H tones.In downstep,each successive H tone in longer downstepping sequences is lower than the preceding one,creating a cumulative staircase pattern.1.PREVIOUS STUDIES ON DOWNSTEP1.1 StewartThe typology of the Twi tone system (1)The pitch of H tones in downstepping sequen-

3、ces in Akan is sensitive to the number of follo-wing downsteps.(2)The pitch of any particular high tone is raised by as many levels as there are downsteps in the subsequent part of the Phrase.(3)The last H tone in the sequence tends to be re-alized at a constant level,its basic pitch.1.2 Schachterso

4、me comments on J.M.Stewarts The typology of the Twi tone system(1)The pitch of the first H tone in Akan is normally phonetically the same regardless of the number of following downsteps.(2)later H tones descend to lower and lower values as the number of downsteps increases.1.3 Studies related to ton

5、al downstep in Chinese Y.Xu proved that anticipatory and carry-over tonal influences co-exist in Chinese,and they differ both in magnitude and in nature.Carry-over effects are mostly assimilatory:the starting F0 of a tone is assimilated to the offset value of a previous tone.Anticipatory effects,on

6、the other hand,are mostly dissimilatory:a low onset value of a tone raises the maximum F0 value of a preceding tone.C.Shih pointed out that the F0 contour of a Chinese sentence is affected by a number of factors,such as declination,downstep and final lowering,etc.X.Huang,Y.Yang and S.Lu reported tha

7、t in downstep in Chinese,the low tone will compress the pitch range of the following syllables down-wards,and the main influence of downstep is on the topline.Their study only analyzed downstep of utterances with 6 syllables,without comparison among utterances of different lengths The experiment rep

8、orted here studied the pitch values of sentences with 2 to 7 HL sequences,i.e.sentences with 4 to 14 syllables,and the aim is to find out whether Chinese sentences display downstep effects across H and L tones.The following questions were addressed:a)Do H tone sequences show downstep effects?b)Do L

9、tone sequences show downstep effects?c)Is the initial H tone scaled higher as the number of downsteps increases?d)Does the final H tone drop to lower values in sentences with more downsteps?2.Methodology2.1 Corpus Design There are four tones in Chinese.Tone 1 is high,Tone 2 rising,Tone 3 low falling

10、,and Tone 4 is a falling one.In order to address the questions of the present experiment,only Tone 1(H)and Tone 3(L)sequences are used here.In the sentences designed,H and L tones alternate on successive syllables,that is,in the pattern of HLHL,HLHLHL In the corpus,each set conta-ined 6 sentences,wi

11、th 2 to 7 HL sequences,i.e.with 4 to 14 syllables in length.(1)Bianxie chugao.(2 HL sequences)To compile the draft.(2)Bianxie gepu chugao.(3 HL sequences)To compile the music draft.(3)Qinshou bianxie gepu chugao.(4 HL sequences)To compile the music draft himself.(4)Kaishi qinshou bianxie gepu chugao

12、.(5 HL sequences)To begin to compile the music draft himself.(5)Jinzao kaishi qinshou bianxie gepu chugao.(6 HL seq-uences)To begin to compile the music draft himself this morning.(6)Zhangwei jinzao kaishi qinshou bianxie gepu chugao.(7 HL sequences)Zhangwei began to compile the music draft himself

13、this morning.2.2 Subjects and Recording The sentences used in this experiment were recorded by eight native speakers of standard Chinese,four males and four females.The test sentences for the experiment were recorded in a sound-treated room at the Phonetics Laboratory in Jinan university,with a shor

14、t practice session before the actual recording.The sentences were presented in random order and was read 3 times by the subject,with the order of each repetition randomized separately.At the recording,the subjects were instructed to read in normal speed,in a natural style,and not to place narrow foc

15、us on any words.2.3 Segmentation of the Signal and Extraction of F0 Acoustic data were segmented and labeled,and F0 was extracted using Praat.The segmenta-tion was first done by a segmenting program and then manually corrected by visual compa-rison of the spectrograms with synchronized waveforms,and

16、 by playback of selected segm-ents for auditory verification.The extracted F0 was manually verified with reference to the cycle in the waveform.The duration of each syllable and the average F0 value of the H tones are calculated,and the lowest F0 value of the L tone is picked.Statistic analysis was

17、done in SPSS.Semitone is used as the unit of pitch,instead of Hertz,and the conversion is done by the following formula:020min12logFStF 3.RESULTS The following Figures graph the mean pitch values of sentences of different lengths for all the eight speakers,with(a)to(f)presenting values for sentences

18、 with 2 to 7 HL sequences respectively.In these graphs,the x-axis displays duration,and the y-axis displays pitch values in semitone.The line segments present the pitch contour,with each segment for one syllable,level one for H tone and low falling one for L tone.3.1 Downstep Effect of H Tone Sequen

19、ces The following figures reveal a prominent gradual lower-ing of H tones strongly resembling downstep throughout the sentences.Table 1 shows the mean pitch values of H tones at different positions for all the eight speakers,and results are listed separately for each sentence leng-th.The first colum

20、n of the table shows the numbers of HL sequences in the sentence,and the second column shows the mean pitch values of H tones at each position,with the ordinal numbers in the second line indicating the position of an H tone in the sentence.The pitch value here is the average of the syllable,rather t

21、han the high point.Also displayed are the significant results from one-factor repeated measures ANOVA.From Table 1 it can be seen that,no matter how long the sentence is,the difference among the pitch values of H tones is always significant,that is,downstep of H tones exists in all the sentences,fro

22、m those with 2 HLs to those with 7 HLs.3.2 Downstep Effect of L Tone Sequences From these figures it can be also seen that there is a consistent lowering of the low points of L tones,regardless of the length of the sentences.Table 2 shows the mean pitch values of the low points of L tones at differe

23、nt positions,and res-ults are also listed separately for each sentence length.Also displayed are the signifificant results from repeated measures ANOVA.From Table 2,it is shown that,similar to the H tones,no matter how long the sentence is,the difference among the pitch values of the low points of L

24、 tones is always significant,that is,downstep of L tones also exists in all the sentences.3.3 The Initial H Tone In Table 1,the first value in each line is the pitch of the initial H tone.If initial H tones were scaled higher as the number of downsteps increases,the first values in each line should

25、rise steadily as we scan them downwards.It can be easily seen that this is the case except for the last one.Result from a repeated measures ANOVA shows that there is significant difference among the pitch values of the initial H tones,F(5,475)=15.71,p 0.001.The initial H tones do shift upwards as th

26、e number of downsteps increases.3.4 The Final H Tone With the effect of downstep,it is expected that as the number of downsteps increases,the final H tone will drop to lower and lower values.When the last values in each line in Table 1 are examined,it is true that there is a continual decline for th

27、em.Result from a repeated measures ANOVA shows that there is also significant difference among the pitch values of the final H tones,F(5,475)=78.7,p 0.001.The final H tones do drop downwards as the number of downsteps increases.3.5 The Final L Tone The last value in each line in Table 2 is that of t

28、he low point of the final L tone.If the last L tones descend to lower values as the number of downsteps increases,the values should decrease steadily.It can be seen that this is the case,but the declined values are very small.Result from a repeated measures ANOVA shows that the difference is margina

29、lly significant,F(5,475)=2.38,p 0.05.So,for sentences with 3 or more than 3 HL sequences,the low points of the final L tones do not descend as the number of downsteps increases.The bottom points of sentences with 3 or more than 3 HL sequences tend to keep constant.4.Discussion4.1 why the initial H t

30、ones in the downstepping sequences shift upwards as the number of downsteps increases This pattern can be attributed to foresight in tone production,that is,the pitch of H tones in down-stepping sequences is sensitive to the number of the following downsteps.When there are more downsteps in a senten

31、ce,the speaker foresees this and will specify the initial H tone at a higher level.From the data of this study it is shown that this is true for sentences with less than 7 HL sequences.4.2 Why when there are 7 HL sequences,the initial H tone will not rise any further.We speculate that this is due to

32、 the limitation of the speakers pitch range.For utterances with no special focus or emotions,the maximum pitch value of the initial H tone is from 14.1 to 15.85 st at a 95%confidence interval,which is obtained from the SPSS analysis of the highest initial H tone,that of sentences with 6 HL sequences

33、.4.3 The quantitative aspects study4.3.1 The Gradient model of downstep Liberman and Pierrehumbert defines downste-ping patterns as a gradual decay toward an abstract reference line,or asymptote.1()nnHd Hrr The method of pitch assignment describes an exponentially decaying curve in which each step d

34、own is proportionally identical to the preceding one in terms of its distance from the reference line r:Later downstep intervals are progressi-vely smaller than earlier ones,and tend to bec-ome vanishingly small as the reference line is approached.This approach could be called a Soft-landing model o

35、f downstep implementation as it describes a curve similar to that of an airc-raft gliding smoothly down to a landing strip.4.3.2 Laniran and Clements expression 11()nnHdHrr4.3.3 Did Chinese speakers apply the Soft-landing algorithm described in（3）we will show the test of data of sentence with 5 HL s

36、equences.In Table 1,the value of the last H tone of the 5-HL sequences is 11.4 st,so we can take that as the value of the reference line r.The value of d is then calculated on the basis of the first H(H1),the second H(H2)and r by,A comparison of the predicted values by Model(3)and the observed value

37、s is given in Table 3.21HrdHr It can be seen that the prediction is largely good,at least better than Linear Model prediction.Similar results can be obtained for the prediction of other HL sequences,so the Soft-landing algorithm can be used in pitch predicting in Chinese.5.Conclusion1.Both H tone se

38、quences and L tone sequences show downstep effects.2.The initial H tone will be specified at higher level for longer sequences when there are 6 or less than 6 HL sequences.3.There are two measures for the implementation of dowstep in Chinese,to raise the initial H tone and to depress the final H one

39、.4.For sentences with 3 or more than 3 HL sequences,the low points of the final L tones do not descend as the number of downsteps increases.5.The Soft-landing model is,to some extent,suitable in pitch predicting in Chinese.Thanks!(a)Sentence with 2 HL sequences(b)Sentence with 3 HL sequences(c)Sente

40、nce with 4 HL sequences(d)Sentence with 5 HL sequences(e)Sentence with 6 HL sequences(f)Sentence with 7 HL sequences返回3返回3.1TABLE I.THE MEANS OF THE AVERAGE PITCH VALUES(IN ST)OF H TONES AT DIFFERENT POSITIONS返回3.1返回3.3返回3.4TABLE II.MEAN PITCH VALUES(IN ST)OF THE LOW POINTS OF L TONES AT DIFFERENT POSITIONS返回3.2返回3.5TABLE III.THE PREDICTED VALUES(IN ST)BY THE SOFT-LANDING MODEL AND THE OBSERVED VALUES FOR 5-HL SEQUENCES返回4.3.3