child-directed phonetic transcriptions

[Brian MacWhinney]

Dear Info-CHILDES,
  Arun Bhalla's concerns about locating phonetic transcriptions of
child-directed speech underscore a more general issue regarding the CHILDES
database.  The database has only a few sets of data in either phonological or
phonetic transcription.  In particular, we have the phonologically
transcribed segments of the Bernstein-Ratner and Cruttenden data for English
and the Levelt/Fikkert and Beers data for Dutch.  In addition, both the
Wilson/Peters and Deuchar data devote attention to phonological detail.
    In principle, audio digitization of the raw data might help people like
Arun, since these digitized audio data could be coded phonological by a
secondary user.  We now have about 300 hours of corpora in digitized audio
format, but only 20 of these hours have transcripts linked to the audio and
none of the digitized corpora have phonetic transcription.  So, this is not
much help at this point.
  Constructing a corpus of phonetic transcriptions is a labor of love and the
few groups that have committed themselves to this difficult work are
understandably possessive about their data.  In addition, the lack of
standardization of computerized IPA fonts inside the larger scheme of Unicode
and the fact that everyone transcribes data slightly differently have made
researchers additionally reluctant to share their phonological transcriptions.
  Despite these barriers, I am optimistic that researchers are coming to
understand the importance of constructing a richer database of phonetic
transcriptions.  If you have computerized or computerizable phonetically
transcribed data that you wish to contribute to CHILDES, we will do
everything possible to help in bringing those data into a format that can be
accessed through the CLAN programs.  In addition, Kim Oller is working on a
translation from CHAT format to his LIPP program, so it should be possible to
derive further analytic benefit from your data by putting them in CHAT format
and also doing LIPP analyses.

--Brian MacWhinney