What is a transcription
A transcription is the orthographic description of a sound file. The
sound is considered as a collection of short sound segments separated by
synchronization breakpoints. Each speaker change mark the beginning of
a new turn. The turns are themselves grouped in homogeneous sections on
different topics.
A typical transcription session
Start a session
- To start a new transcription, choose menu File/New trans. You will
be asked for a signal to transcribe. This choice can be changed later with
menu File/Open audio file...
- Save your work often with menu File/Save. The first time you save a
new transcription, you will be proposed the same name as the signal (but
with a different extension, depending on the chosen format -you should
not type in manually this extension).
- If you want to work on an existing transcription, open it with menu
File/Open trans... Transcriber will look for signal sharing the same name
as the transcription; if it fails, you will be asked to give location of
the signal with a dialog box.
If you save the configuration with menu Options/Save configuration before
leaving, both sound and transcription files will be re-opened automatically
the next time you launch Transcriber.
Listen to the signal
Signal can be manipulated with mouse (scrollbars, buttons), menus, or
keyboard shortcuts. Use 'Tab' to start/pause playback.
First pass: rough segmentation
- Choose a resolution of 30 seconds or less for the signal
- Start playback at the beginning of the file (or at the beginning of
the untranscribed part of the signal).
- During playback, insert a new breakpoint with the 'Enter' key each
time you ear short pauses in the speech or changes in background. A new
line with a point will appear in the text, and a bar will appear under
the signal a the position of the cursor.
- After a few minutes of rough segmentation, you should get a succession
of segment of 10 to 30 seconds.
Don't forget to save your work often (menu File/Save).
Second pass: transcription
- Choose a resolution of 10 seconds or less for the signal
- Go back to the first untranscribed segment by clicking on the first
line of the text or the beginning of the signal.
- Listen to the segment (menu Signal/Replay segment)
- If necessary, adjust the segment boundaries by clicking with button-2
on the segment boundary and draging the mouse
- If a new speaker starts speaking at the beginning of the current segment,
mark it as a new turn (menu Segmentation/Change segment type/Turn, or its
shortcut). You will be asked the speaker name, but you can keep the default
(New speaker) and modify it later if you want.
- Type in the transcription; it will automatically appear in the text
editor and in the segmentation under the signal
- Process in the same way all remaining segments
Third pass: verification
- Go back and listen to the signal while reading the transcription. Start
and pause playback as often as you need (use the 'Tab' key). It is possible
to move forward/backward during playback.
- Structure the document at the global level by marking new sections
(menu Segmentation/Change segment type/Section, or its shortcut). Section
types are 'report' , 'filler' or 'nontrans'. You can attribute an arbitrary
topic to the section.
- Transcribed text can be modified in the text window. Speaker names
and topic can be modified from the menu Segmentation/Edit attributes...
or directly clicking on the button in the text window