Master's Thesis Project: KaRS

The Kiel Corpus of Read Speech as a Resource for Speech Synthesis

Thesis

postscript [3.5MB]
zipped postscript [800KB]
postscript (2 pages on one) [3.5MB]
PDF [700KB]
PDF (2 pages on one) [650KB]

Added Features

Files with features added to the Kiel Corpus of Read Speech (KCoRS). Due to copyright restrictions, you need a copy of the KCoRS in order to obtain these files. Please send me an e-mail if you are interested.

• sentence types
• part-of-speech, chunk phrases, and word frequency
  • automatically derived (with TnT and chunk tagger)
  • manually corrected
• chunk phrases and grammatical functions
  • automatically derived (with SCHUG)
  • manually corrected
• syllables
  • canonic syllables with lexical stress
  • realised syllables

Classification and Regression Trees

• symbolic prosody prediction
  • prosodic boundary location: kko/k61    rtd/k62
  • accentuation level: kko/k61    rtd/k62
  • accent
    • accent location: kko/k61    rtd/k62
    • accent type: kko/k61    rtd/k62
  • phrase-final intonation contour: kko/k61    rtd/k62
• prediction of postlexical phonological processes
  • type of change
    • Symbolic: kko/k61    rtd/k62
    • Direct: kko/k61    rtd/k62
  • replacement rule
    • Symbolic: kko/k61    rtd/k62
    • Direct: kko/k61    rtd/k62
• prediction of acoustic parameters (z-scores)
  • duration
    • Symbolic: kko/k61    rtd/k62
    • Direct: kko/k61    rtd/k62
  • median F0
    • Symbolic: kko/k61    rtd/k62
    • Direct: kko/k61    rtd/k62
  • last F0
    • Symbolic: kko/k61    rtd/k62
    • Direct: kko/k61    rtd/k62

Sound Files for Perceptual Evaluation

male voice

sentence	CopySynth	Symbolic	Direct	MARY
be006
be038
be074
cn015
e026
e040
e042
ko029
ko039
ko049
mr007
mr016
mr018
mr040
mr088
s041
s072
s1017
s1040
tk010

female voice

sentence	CopySynth	Symbolic	Direct	MARY
be006
be038
be074
cn015
e026
e040
e042
ko029
ko039
ko049
mr007
mr016
mr018
mr040
mr088
s041
s072
s1017
s1040
tk010

Any questions, comments, and suggestions welcome!