next up previous contents index
Next: Manual segmentation Up: Concerning the segmentation and Previous: Use of segmented speech

The CRIL conventions


The CRIL (Computer Representation of Individual Languages) conventions have been defined and proposed by a working group at the 1989 Kiel Convention of the International Phonetic Association . The conventions consist of two parts, as follows:

  1. A mapping of IPA  categories to standard IPA numbers and IPA  symbol names.gif
  2. Three distinct levels of representation for natural speech production.

The first component was introduced to enable broader use and dissemination of the descriptive IPA  categories: i.e. the IPA  symbols as well as the IPA  diacritics needed for the narrow transcription  of normal and defective utterances.

The second component of CRIL is devoted to a standardised representation of natural speech productions and introduces three systematically distinct levels for specifying what could be called the text of a spoken utterance. These levels are as follows:

  1. Orthographic level
    This level contains the orthographic representation of the spoken text.
  2. Phonetic level
    This level specifies the phonetic form of a given word in its full (unreduced) segmental  form. This form would only appear when the word is spoken in isolation: i.e. it is the citation form.  
  3. Narrow Phonetic level
    This level gives the narrow phonetic  transcription  of the words that were actually spoken. It is only on this level that phonetic categories can be directly related to the speech signal itself.

The CRIL conventions have been shown to work very well for the characterisation of speech data in the German PHONDAT and VERBMOBIL-PHONDAT  corpora. However, the SAM  conventions described below offer the opportunity to represent more detail.  

EAGLES SWLG SoftEdition, May 1997. Get the book...