Decision Tree learning on Phonetic Error

A phone is the smallest identifiable unit found in a stream of speech. One same symbol can be pronounced differently in different languages. For example, character “a” pronounced as “[ei]” whereas it is pronounced similar as “ah” in English. Therefore, it is essential to analyse the underlining patter of phonetic error made by the second language learners in this corpus. Decision tree is used to analyse the pattern of each syllable error produced in this corpus.  However, due to the space limitation, only the syllables with high error rate  will be analysed. The decision trees of top 10 syllables have be generated.  The following figures show 2 sample decision trees generated for syllable “d” and syllable “ch” by using a Python script.

Figure 14 Decision Tree  of syllable “d”

Figure 15 Decision Tree of syllable “ch”

* There are total 10 decision trees have been generated by the Python script.