Contents

• List of Figures
• Introduction
• Review of the Literature
  • Topic segmentation
    • TextTiling and Marti Hearst's work
    • Research extending and improving, and using TextTiling
    • Other approaches
  • Spoken dialogue and structure
    • Theories of structure
  
  
• Software Implementation
  • Outline of functionality
  • Architecture
  • Implementation details
    • Ancillary modules
    • Parsing the data
    • Output
    • Visualisation
  • Summary of usage
  • Example session
    • Calculating similarities and detecting breaks
    • Creating annotated transcripts in HTML
    • Plotting graphs
  • Bugs and inefficiencies
  
  
• Experiments
  • Outline
  • Methodology
    • Hand-annotated data collection
    • System parameter experiments
  
  
• Results and Evaluation
  • Calibration data
  • Evaluations against human markup
    • Imprecision
    • Misses
    • Annotator normalisation
    • Sparseness agreement
    • Annotator differences
  • Parameter tuning
  
  
• Conclusions
  • Observations
    • Implications of false positives
    • Inter-annotator agreement
  • Further work
  
  
• Bibliography