TGA: An Online Tool for Time Group Analysis

• Dafydd Gibbon
• http://wwwhomes.uni-bielefeld.de/gibbon/TGA

TGA specifications

• Requirements specification
• Design and implementation
• Input parameters
• Outputs

TGA Requirements specification (1)

• Annotation mining: the extraction of information from annotations, e.g. Praat TextGrids.
• In speech technology, annotated data are generally mined (semi-)automatically and efficiently.
• In phonetics, manual or semi-manual mining is common but inefficient:
  • copying Praat information into a spreadsheet
  • defining functions sich as nPVI in the spreadsheet
  • calculating and generating graphics
• In phonetics there is a need for faster and more consistent mining of larger numbers of TextGrid files, without necessarily working with programming experts

TGA Requirements specification (2)

• The Time Group Analyzer (TGA) is designed to support phoneticians by automatizing a wide range of relevant computational tasks:
  • duration extraction from TextGrids to table format,
  • basic descriptive statistics, slope, nPVI …,
  • novel visualisations of timing structure:
    • global acceleration/deceleration patterns
    • local acceleration/deceleration (trochaic/iambic, shorter/longer) Duration Difference Tokens (DDTs) and DDT sequences, for study of rhythm
    • Time Trees, for comparison of timing with grammatical structure
    • Wagner Quadrant plots
    • Box plots of unit durations

Design and implementation (1)

• Software Development Environment: Python 2.7, CGI
• Input:
  • Praat TextGrid (long or short),
  • CSV (Character Separated Values, with various separator chars).
• Output:
  • HTML with text, syllable props, pausal group stats, Diff Tokens, Time Trees,
  • CSV for further processing.

Design and implementation (2)

TGA Input parameters

• Input form
• Input control parameter choices
• Time Group duration difference parameters
• Output parameter choices

TGA input form - screenshot

TGA input form: parameter choices

• Input control parameter choices:
  • Textgrid tier name selection (e.g. 'Syllables', 'syllable', 'syll' - the tier can also be other items than syllables)
  • Pause symbol selection (e.g. '_', 'p', 'sil') for segmenting into interpausal groups
• Time Group duration difference parameters:
  • Local TG threshold: sets the minimal difference (in ms) which counts as a difference; any difference below this threshold counts as equal duration
  • Local TG pattern symbols: select the symbols used for longer, shorter and equal duration difference relations ('duration difference n-grams')
  • Global threshold range: ignore this one; not used currently
  • Minimum TG length in syllables (e.g. 2, 3)

Output parameter choices

• Select various outputs, or all outputs (default):

  • Text extracted from labels
  • General information about TG elements
    • descriptive statistics, nPVI, regression slope and intercept
  • Details about individual interpausal groups:
    • descriptive statistics
    • visualisation
    • Duration Difference Token (DDT) sequences
    • Time Trees (TT) types
  • DDT n-grams
  • TT types
  • Conversion of input TextGrid to Character Separated Value (CSV) format

TGA output

• Text extraction
• Statistics
• Time Group visualisation
• Duration Difference Token sequences and n-grams
• Time Trees
• CSV formatted output of TextGrid input

TGA output: text extraction

• automatic transcription text extraction is convenient for work styles in which transcription and annotation are simultaneous

TGA output: syllable duration

TGA output: descriptive statistics

TGA output: Wagner Quadrants

TGA output: duration bars and DDTs

TGA output: DDTs and Time Trees (TTs)

TGA output: DDT n-gram distribution

TGA output: Time Tree (parenthesis notation)

Applications in timing analysis

• Comparison of different TGA measures

• Planned: box plots of tiers with few text types

Comparison of measures in Aix-MARSEC genres

• Visual comparison of different properties of duration sequences
  • for example: standard deviation and nPVI
  • interesting: duration sequence slope (scaled to fit graph)

Planned: box plots for tiers with few text types

• box+whisker plot, error bar (left), duration distribution (right)

Further information on TGA web site:

• Demos with English, Mandarin and Tem (Niger-Congo) TextGrids
• Links on demo pages to relevant papers and presentations
• TGA Home Page Link

Summary

• Introduction
• Selection of annotation software
• Corpus development methodology
• Momel and INTSINT
• SPPAS
• Time Group Analyzer
• Conclusion and references