Amplitude and Cross-Speaker Vowel Space Similarity

This Shiny interactive enables exploration of judgements of across-speaker vowel space similarity and their dependence on relative amplitude, measured by z-scoring amplitiude for each speaker.

The visualisations we generate have two rows:

  1. a high amplitude interval for a speaker and the three closest vowel spaces in the corpus.
  2. a low amplitude interval for the same speaker, and the three closest vowel spaces in the corpus.

The controls are:

  • Interval length: whether the corpus is divided into 60 or 240 second intervals.
  • Speaker: the speaker we will select intervals from.
  • High Amplitude Interval: a list of intervals from the selected speaker with high (>0.3) mean amplitude values.
  • Low Amplitude Interval: a list of intervals from the selected speaker with low (<0.3) mean amplitude values.
  • Normalisation method: method for normalising data, with options:
    • 'Raw Hz, whole recording': mean value for each monophthong in Hz across whole recording.
    • 'Lobanov 2.0, within interval': the Lobanov 2.0 normalisation method (see Brand et al. 2021) applied within each interval.
    • 'Lobanov 2.0, whole recording': same as above, but calculated across whole recording.
  • Scale distances: allows for distances between monopthongs in each vowel space to be scaled so that vowel distances are relative to the distribution of distances between vowels in the corpus. i.e. the distance between DRESS in one vowel space and DRESS in another will be relative to the distribution of distances between DRESS vowels in the corpus as a whole. This means that vowels with a lot of 'room to move' in the vowel space are less determinative of vowel space similarity.

The PC1 score for each interval is also given. In our PCA analysis, PC1 patterns with amplitude.

This interactive was designed for the purpose of the paper 'The overlooked effect of amplitude on within-speaker vowel variation' (Linguistics Vanguard, 2023, https://doi.org/10.1515/lingvan-2022-0086).

Code for this interactive is available at: https://github.com/JoshuaWilsonBlack/vs_similarity_shiny