A Requirement of Texts for Evaluation of Rhythm in English Speech by Learners

In objective evaluation of rhythm in English speech uttered by learners, the prediction accuracy is greatly affected by the properties of texts given to the subject to speak. In order to construct an objective evaluation model for more efficient prediction of the subjective evaluation scores given by evaluators, we examined a requirement of texts for evaluation of rhythm in English speech. A model was constructed based on acoustic differences in speech between learners and native speakers. Repetition of stressed syllables alternating with unstressed syllables helps listeners perceive rhythm in English. Focusing on the property of salient stressed syllables in the repetition, the number of stressed syllables included in a text was treated as an element of a requirement. As a result of prediction experiments, it became clear that a text including three stressed syllables is an effective requirement of texts for evaluation of rhythm in English.