Generate rules from Decision Tree of Tonal Error
The path from root to leaf in a decision tree model represents the set of classification rules. The classification rule is able to provide a summery of the entire model and serves as the guideline to classify the new incoming data. The classification rules generated from the five decision tree models were shown in the following figures.
- Rules for Tone 1
No. |
Error Pattern |
Probability (No.of samples) |
Examples |
T1a | [+1] –> [+3] / [+3*] [ ] [-4*](Previous said 3 is true and next said 4 is false) | 0.691 | 点心 dessert
dian3 xin1 –> dian3 xin3 |
T1b | [+1] –> [+4] / [+Uttinitial][ ]Tone 1 character is the first character in the sentence | 0.553 | 辛苦 exhausting
xin1 ku3 –> xin4 ku3 |
T1c | [+1] –> [+4] / [+3*] [ ] [+1*] | 0.543 | 电芯盒 batteries box
dian4 xin1 he2 –> dian3 xin4 he1
|
T1d | [+1] –> [+4] / [-3*] [+FinalVowel] | 0.531 | 汽车 car
qi4 che1 –> qi4 che4
|
T1e | [+1] –> [+3] / [+Uttinitial][+1*] | 0.518 | 关注 attention
guan1 zhu4 –> guan3 zhu1 |
T1f | [+1] –> [+2] / [-Uttinitial][ ][+Short] | 0.478 | 不吃饭 do not eat
bu4 chi1 fan4 –> bu4 chi2 fan4 |
T1h | [+1] –> [+4] / [ ] [+4*] | 0.425 | 轰隆 bang
hong1 long2 –> hong4 long4 |
T1g | [+1] –> [+4]: Default (all context other than the rules listed above) | 0.423 |
- Rules for Tone 2
No. | Error Pattern | Probability (No.of samples) | Examples |
T2a | [+2] –> [+1] / [ ] [+NasalFinal] [ ] | 0.579 | 严肃 serious
yan2 su4 –> yan1 su4 |
T2b | [+2] –> [+4] / [ ] [+InitSibil] [ ]*Initial Sibilant: z, c, s | 0.527 | 责备 blame
ze2 bei4 –> ze4 bei4 |
T2c | [+2] –> [+3] / [ ] [+SonorantInital] [+1*] | 0.518 | 不如意 disappointment
ru2 yi4 –> ru3 yi1
|
T2d | [+2] –> [+1] / [+3*] ____ | 0.473 | 冰河 glacial
bing1 he2 –> bing3 he1 |
T2e | [+2] –> [+3] / [+UttInitial ] [ ]+SonorantInitial+NasalFinal
|
0.462 | 能力 ability
neng2 li4 –> neng3 li4 |
T2f | [+2] –> [+4] / [ ] [+SonorantInitial] [ ] | 0.453 | 开营式 camp opening ceremony
kai1 ying2 shi4 –> kai1 ying4 shi4
|
T2g | [+2] –> [+1] / [ ] [+NasalFinal ]+SonorantInitial | 0.42 | 中文 Chinesezhong1 wen2 –> zhong1 wen1 |
[+2] –> [+1]: Default, case other than T2b | 0.434 | 水渠 drains
shui3 qu2 –> shui3 qu1 |
|
T2h | [+2] –> [+4] / [ ] [+Diphong] [ ] | 0.416 | 挨打 take a beatingai2 da3 à ai4 da3 |
T2i | [+2] –> [+1] / [ ____ ] [+4*] | 0.392 | 河川 river
he2 chuan1 –> he1 chuan4 |
- Rules for Tone 3
No. | Error Pattern | Probability (No.of samples) | Examples |
T3a | [+3] –> [+2] / [-3*][+UttFinal] | 0.639 | 新领域 new areas
xin1 ling3 yu4 –> xin1 ling2 yu4
|
T3b | [+3] –> [+2] / [ ] [+SonorantInitial] [ ] | 0.559 | 电影院 cinema
dian4 ying3 yuan4 –> dian4 ying2 yuan4
|
T3c | [+3] –> [+1] / [-3*] [-FinalVowel] | 0.474 | 凶猛 firece
xiong1 meng3 –> xiong1 meng1
|
T3d | [+3] –> [+2] / [ ] [+Short] [ ] | 0.461 | 养老院 nursing home
yang2 lao3 yuan4 –> yang2 lao2 yuan4
|
T3e | [+3] –> [+1] / [ ] [+NasalFinal ]+InitialVoiced | 0.459 | 充满 full
chong1 man3 –> chong1 man1 |
T3f | [+3] –> [+2] / [ ] [+NasalFinal ] [ ]-InitialVoiced | 0.449 | 核反应 nuclear reaction
he2 fan3 ying4 –> he2 fan2 ying4 |
T3g | [+3] –> [+1] / [+3*] [+FinalVowel][-InitialVowel] | 0.426 | 练武术 practicing martial arts
lian4 wu3 shu4 –> lian3 wu1 shu4 |
T3h | [+3] –> [+2] / [ ] [+Diphthong] [ ] | 0.412 | 偶然 accidental
ou3 ran2 –> ou2 ran2
|
T3i | [+3] –> [+2] / [+3*] [+FinalVowel] [+InitialVowel] | 大企鹅 big penguin
da4 qi3 e2 –> da3 qi2 e2 |
|
T3j | [+3] –> [+1]: Default any case other than T3h | 0.348 | 以为 that
yi3 wei2 –> yi1 wei2 |
- Rules for Tone 4
No. | Error Pattern | Probability (No.of samples) | Examples |
T4a | [+4] –> [+2] / [ ] [+SonorantInitial]+UttFinal | 0.544 | 八日 the eighth (of the month)
ba1 ri4 –> ba1 ri2
|
T4b | [+4] –> [+1]: Default any case other than T4a | 0.505 | 大人 adult
da4 ren2 –> da1 ren2 |
T4c | [+4] –> [+1] / [ ] [-DiphFalling] [-1*] | 0.493 | |
T4d | [+4] –> [+3]: Default any case other than T4c | 0.48 | |
T4e | [+4] –> [+2] / [ ] [+SonorantInitial] [-1*] | 0.468 | 易容 disguise
yi4 rong2 –> yi2 rong2
|
T4f | [+4] –> [+1] / [+3*] [+SonorantInitial] [ ] | 0.436 | 容易 easy
rong2 yi4 –> rong3 yi1
|
T4g | [+4] –> [+1] / [+UttInitial ]+SonorantInitial | 0.434 | 利息 interest
li4 xi1 –> li1 xi1
|
T4h | [+4] –> [+3]: Any case other than T4d | 0.421 |
- Rules for Tone 5
No. | Error Pattern | Probability (No.of samples) | Examples |
T5a | [+5] –> [+2] / [ ] [+NasalInitial] [ ] | 0.566 | 什么 what
shen2 me5 –> shen2 me2 |
T5b | [+5] –> [+4] / [+3*] [+FinalVowel] [ ] | 0.5 | 案子 case
an4 zi5 –> an3 zi4 |
T5c | [+5] –> [+4] / [ ] [+FinalVowel ] [ ]+InitialRetroflex | 0.434 | 叔叔 uncle
shu1 shu5 –> shu1 shu4 |
T5d | [+5] –> [+1]: Any case other than T5f | 0.416 | 婶婶 aunt
shen3 shen5 –> shen3 shen1 |
T5e | [+5] –> [+4] / [-1*] [+FinalVowel] [ ] | 0.408 | 胖子 fatty
pang4 zi5 –> pang4 zi4 |
T5f | [+5] –> [+2] / [+FinalVowel] [___] [ ] | 0.361 | 瘦子 thin
shou4 zi5 –> shou4 zi2 |
T5g | [+5] –> [+3] / [ ] [+FinalVowel] [ ]+DiphFalling | 0.325 | |
T5h | [+5] –> [+4] / [+1*] [+FinalVowel] [ ] | 0.311 | 矮子 short person
ai3 zi5 –> ai1 zi4 |
T5i | [+5] –> [+1] / [ ] [+FinalVowel ] [ ]+RepeatSyllable | 0.3 | 爸爸 father
ba4 ba5 –> ba4 ba1 |
* Diphthong: ai, ei, ao, ou