Week 9 (March 17 – 21, 2014)

Decision Tree Learning on Tonal Error

Mandarin is a tonal language therefore tone is essential in distinguishing Mandarin syllables. The data collected in the existing corpus were from non-tonal language speakers. Thus, tonal error occurs significantly. Tonal error is defined as mispronunciation of tone due to substitution of other tone. The tonal errors were analyzed by generating decision tree for each of the tone (i.e. Tone 1,2,3,4,5). The following figures show the five decision trees generated by using a Python Script.

1

Figure 9 Decision Tree of Tone 1

5

Figure 10 Decision Tree of Tone 2

2

Figure 11 Decision Tree of Tone 3

3

Figure 12 Decision Tree of Tone 4

4

Figure 13 Decision Tree of Tone 5