AI4LAM Speech-to-Text WG, May 27, on transcript metadata

[Owen King]

The AI4LAM Speech-to-Text Working Group invites you to its next meeting on Tuesday, May 27 at 09:00 US-Pacific | 12:00 US-Eastern | 17:00 UK | 18:00 Central Europe | 03:00 +1 Canberra.

Our topic will be metadata for transcripts.  Many libraries, archives, and museums create and use transcripts associated with audiovisual resources.  Transcripts come from various sources, and they vary in quality and other characteristics.  For example, if a given transcript was created by Whisper, one would like to know:  With which model?  Using what parameter values?  Has it been corrected by a human?  According to what conventions?  Transcript metadata can answer such questions.

We will discuss recent attempts to organize transcript metadata, including FADGI's Guidelines for Embedding Metadata in WebVTT Files, available here: https://www.digitizationguidelines.gov/guidelines/accessibilty_WebVTT.html 

We will also discuss recent work in this group to define a more expansive set of elements, called Transcript Provenance Metadata Elements (TPME).  The current draft specification, about which we're currently seeking feedback, is available here:   https://docs.google.com/document/d/10hEbp_RkOeSm5uorTlI_QmyE0sD1G5OAkWUfHICf8W0 

Please join us!

Agenda and running notes: https://docs.google.com/document/d/1lUI1l_cfJ-hM7ZXgfITyjcevUxFfc0C6HRzhc_Ui8bU 

Zoom: https://stanford.zoom.us/j/99320941121?pwd=AafIBuc5maw5mcsiHYrcW7uSQmB6t5.1&from=addon 

Cheers,

Owen (on behalf of the Speech-To-Text WG conveners)

Owen King (he/him)

Metadata Operations Specialist

E: owen...@wgbh.org

One Guest Street, Boston, MA 02135