Using covariances to understand tonal coarticulation in Mandarin

In tonal languages such as Mandarin Chinese, the meaning of a word depends on the pitch variation of the tone. Since tones are often not pronounced in isolation, but rather concatenated, neighboring tones effect each other. This gives rise to tonal coarticulation. In this thesis, we will explore if, given two concatenated tones of the Mandarin word “ma”, it is possible to predict the following tone on the basis of the coarticulation effect present in the first tone, and vice versa. The phonetic data used for this exploration hold a certain intrinsic smoothness that points naturally towards the functional data analysis domain as a tool to study them. Therefore, we will be using multiple functional data analysis techniques. We will start with k-means clustering on the raw data with the Euclidean distance and the Manhattan distance. Afterwards, we will study the effect on tone duration, for which we will be using duration analysis. Furthermore, previous research indicates that the...Show moreIn tonal languages such as Mandarin Chinese, the meaning of a word depends on the pitch variation of the tone. Since tones are often not pronounced in isolation, but rather concatenated, neighboring tones effect each other. This gives rise to tonal coarticulation. In this thesis, we will explore if, given two concatenated tones of the Mandarin word “ma”, it is possible to predict the following tone on the basis of the coarticulation effect present in the first tone, and vice versa. The phonetic data used for this exploration hold a certain intrinsic smoothness that points naturally towards the functional data analysis domain as a tool to study them. Therefore, we will be using multiple functional data analysis techniques. We will start with k-means clustering on the raw data with the Euclidean distance and the Manhattan distance. Afterwards, we will study the effect on tone duration, for which we will be using duration analysis. Furthermore, previous research indicates that the coarticulation effect lies at the level of covariances. Hence, we will also be clustering functional covariances. In the last section, indications of the results will be discussed and suggestions will be made for further research. Lastly, plots obtained from the analyses are shown in the appendixShow less

Leiden University Student Repository

Documents

In Collections

Using covariances to understand tonal coarticulation in Mandarin