Analysis of Chinese Syllable Durations in Running Speech of Japanese L2 Learners

Yue Sun, Shudon Hsiao, Yoshinori Sagisaka, Jinsong ZhangInterspeech 2016

TL;DR

This paper investigates how Japanese learners of Mandarin Chinese, as a second language, produce syllable durations differently from native speakers, particularly in running speech. The findings reveal that while native speakers adapt syllable durations dynamically based on tone and context, L2 learners tend to produce more uniform durations, especially for certain tones, suggesting a less nuanced control of prosody.

What Problem Does This Paper Address?

Learning to produce Mandarin Chinese tones accurately is a well-known challenge for second language (L2) learners, especially those whose native languages are non-tonal. While much research has focused on the fundamental frequency (F0) contours of tones, the role of syllable duration in prosody generation and tone realization is also crucial. This paper addresses the gap in understanding how L2 learners, specifically Japanese speakers learning Mandarin, manage syllable durations in running speech compared to native speakers. The authors aim to uncover how tone type influences syllable duration and whether L2 learners exhibit the same dynamic adjustments as native speakers, which is essential for natural-sounding speech.

Methods

The study employed a comparative analysis of syllable durations in running speech. The researchers collected speech data from three groups: native Mandarin Chinese speakers, intermediate Japanese L2 learners, and beginner Japanese L2 learners. All speech samples were manually annotated to label syllables and their corresponding tones. Durational information was extracted using the Praat software.

To account for variations in speaking rate across individuals and groups, syllable durations were normalized. This normalization allowed for a comparison of the variations in syllable durations rather than their absolute lengths in milliseconds. Statistical analyses, including ANOVA and Tukey's HSD tests, were used to determine the significance of differences in mean syllable durations and their variations between tone types and speaker groups. The analysis also considered the effect of the neutral tone (TO) by comparing results with and without its inclusion.

Datasets

The study utilized a speech corpus collected at Beijing Language and Culture University (BLCU) and Waseda University, Japan. This corpus included:

  • Native Mandarin Chinese speakers: 12 participants (6 female, 6 male) from Beijing, with a high proficiency in Mandarin.
  • Intermediate Japanese L2 learners: 19 participants (9 female, 10 male) studying Mandarin at BLCU.
  • Beginner Japanese L2 learners: 10 participants (7 female, 3 male) studying Mandarin at Waseda University for one year.

The corpus consisted of approximately 50 spoken language sentences per speaker, averaging 9.6 syllables per sentence.

Key Insights and Ideas

  • Native speakers are dynamic: Native Mandarin speakers show significant variations in syllable durations across different tones in running speech, adapting them to the context. They tend to produce tones 1 and 2 with longer, more stable durations, while tones 3 and 4 are shorter and more flexible.
  • L2 learners are more uniform: Japanese L2 learners, in contrast, exhibit less variation in syllable durations across tones, particularly for tones 3 and 4. This suggests they might be producing these tones in a more "isolated syllable" fashion rather than integrating them dynamically into running speech.
  • Neutral tone's impact: The neutral tone (TO), being inherently shorter, significantly increases the overall variation in syllable durations for L2 learners. When TO is excluded, the variations decrease, but the fundamental differences between native and L2 speakers remain.
  • Tone 3 and 4 are key differentiators: The most pronounced differences in duration variation between native speakers and L2 learners are observed for tones 3 and 4. Native speakers show much larger variations for these tones, especially tone 3, which is known for its complex realization in running speech.
  • Proficiency matters, but not entirely: While intermediate learners show some differences from beginners (e.g., slightly larger variations), the core distinction between native speakers and L2 learners persists across proficiency levels.

Key Results

  • Mean Syllable Durations: Native speakers' mean syllable durations for tones followed the order: TO, T3, T4, T2, T1. Intermediate learners showed TO, T3, T2, T4, T1, and beginners TO, T4, T3, T2, T1. A notable difference was the longer duration of T4 for beginners compared to natives.
  • Variations in Normalized Syllable Durations:
    • Native speakers consistently showed the largest variations in normalized syllable durations across tones.
  • Japanese L2 learners (both intermediate and beginner) showed significantly smaller variations compared to native speakers.
  • The variations for L2 learners decreased when the neutral tone (TO) was excluded, but the differences between native and L2 speakers remained significant.
  • For lexical tones, native speakers had larger variations for T3 and T4 than L2 learners. L2 learners showed smaller variations for T3 and T4, making them more similar to T1 and T2.
  • Statistical Significance: Differences in syllable duration variations between native speakers and both groups of L2 learners were statistically significant for most comparisons, especially for tones 3 and 4.

Limitations and Future Work

Limitations: * The study focused solely on syllable duration and did not incorporate F0 contours or intonation effects, which are also critical for prosody. * The analysis of tone 3 sandhi (where T3 can change to T2) was simplified by merging these instances with regular T2, which might mask some nuances.

Future Work: * Combine syllable duration analysis with F0 contour analysis to gain a more comprehensive understanding of tone generation in L2 learners. * Investigate the effects of sentence position and intonation on syllable durations for L2 learners. * Explore the underlying phonetic mechanisms and cognitive processes that lead to these durational differences. * Use these findings to develop more targeted pedagogical strategies for teaching Mandarin Chinese tones to L2 learners, particularly for running speech.

Summary updated

BibTeX
@inproceedings{sun2016analysis,
  author = {Yue Sun and Shudon Hsiao and Yoshinori Sagisaka and Jinsong Zhang},
  title = {{Analysis of Chinese Syllable Durations in Running Speech of Japanese L2 Learners}},
  booktitle = {Interspeech},
  year = {2016},
}
Figure