Transcriber User Manual (Transcriber 1.4.4)

User manual of Transcriber 1.4.1

6. Edition

7. Episode

8. Segmentation

9. Speech turns and speakers

10. Sections and topics

11. Background noise

12. Events and comments

13. Keyboard shortcuts

14. Options and interface configuration

15. Other features



1. Startup



1.1. Working environment



The software is meant to allow the transcription of long duration audio signals with standard hardware. For instance, several one hour long recordings (sampled at 16 kHz on 16 bits, mono) were already successfully transcribed with a PC/Pentium 200MHz under Linux. Using headphones is strongly recommended for a quality transcription, but loudspeakers may be useful for checking with other people.

1.2. Starting the tool


When Transcriber is correctly installed, it can be started with the command:

or, depending on the configuration, by clicking a given menu or icon. A dialog box then proposes to select a sound file to create a new transcription associated with this signal, or a transcription file to display or modify it. A short example of signal with an associated transcription are available in the demo/ directory, distributed with the software.

When the working configuration has been saved during a previous session, the current signal and transcription files are automatically opened if they are still available on the disk.

It is also possible to start the tool and give it the signal and/or transcription file as arguments on the command line. If both files are in the same directory, have the same name (except for the extension) and if the signal file has a standard extension (like .wav or .snd), the name of the signal file can be omitted, as it will be automatically opened:

1.3. Initial configuration


After the first use of the tool, it is recommended to open the [Options]/[General...] menu and to fill in:

and more signal specific options, in the menu [Options]/[Audio file...] :

The options should then be saved with the menu [Options]/[Save configuration].

2. Windows


2.1 Menu bar


2.2. Transcription


In the upper side of the interface, the orthographic transcription of the signal can be typed in a text editor.

Each segmentation begins at time t=0; there is at least a section, a turn and a segment on the first 3 lines of the transcription. The line containing the cursor is highlighted. There is a scrollbar to the right.

2.3. Button bar


In the middle, a button bar (which can be hidden) gives shortcuts for signal playback or for information display and shows the current filenames.

Available commands:

2.4. Signal and segmentation


Signal and segmentation appear in the lower half; there is also a scrollbar for choosing the part of the signal to display and another smaller scrollbar (which can be hidden with a contextual menu) for resolution control.

Under the signal, the segmentations (orthographic one, in speech turns, in sections, in background conditions) is displayed and synchronized with the signal; the current segment in the orthographic transcription is highlighted.

A second view of the signal in a different resolution can be displayed. Current cursor position in the signal is shown with a vertical red bar. Both cursors (in the text editor and in the signal) are always synchronized so that the current segment in the text is the same as the current segment under the signal; as soon as one cursor enter another segment, the other one is automatically moved in order to appear in the same segment.

On the signal, mouse controls are the following:

On the segmentation lines, mouse controls are:

3. Transcription file


3.1. Create a new transcription


Choose menu [File]/[New trans]. A dialog window ask for the sound file to be transcribed. This can be changed later.

3.2. Read a transcription


With [File]/[Open trans...], you can read transcription files produced by Transcriber, and transcription or segmentation files in some standard formats identified by the filename suffix:

The signal file associated with the transcription is then searched for; if it is not automatically found, the user is asked to look for it.

3.4. Save the transcription


The transcription can be saved with menu [File]/[Save] if modifications occurred since creation or last save. The first time, transcription filename defaults to signal filename with .trs extension. It can be saved later under another name with [File]/[Save as...].

3.5. Export to another format


The transcription can be written into non-native transcription formats (.typ, .stm or .html) with menu [File]/[Export], but these formats are not as complete as the native one, and information can be lost in the process. However .html format makes printing the transcription possible using any browser.

3.6. Revert to previous version


Command [File]/[Revert] will undo all changes in memory and load lastly saved version on disk.

3.7. Automatic backup


An automatic backup is performed within a fixed time after a modification (if the transcription has not been saved in between). It is saved in the same directory as the transcription with the same name surrounded with # character (for example, /data/transcription/#radio.trs#). This file will be removed the next time the transcription is regularly saved. Time interval is given within window [Options]/[General...]; if it is zero, no automatic backup occurs.

3.8. Backup file


When a previously existing transcription is read on the disk, then modified and saved, the existing version is kept in a file with the same name as the transcription followed by a string given by the user in the window [Options]/[General...] (for example, ~ will /data/transcription/radio.trs~). If the field is empty, the previous file is not kept.

3.9. File encoding


Since Transcriber 1.4.1 under Tcl/Tk 8.1.1 or higher, the encoding for file output can be chosen in [Options]/[General...]. It defaults to ISO-8859-1 (Latin1) but can be changed to other widespread 8 bits encodings (ISO-8859-*, EUC-JP...) but also to Unicode (saved as UTF-8, which is backward compatible with 7 bits US-ASCII). If Latin1 does not work and you are not sure what to use, try UTF-8 as it supersedes the other ones.
Upon reading, the encoding used is the one defined in the file header. For transcriptions missing this information (like the ones produced with earlier versions of Transcriber or Tcl/Tk), or when the encoding is not ASCII-compatible, the current encoding is used.

4. Signal file


4.1. Open a signal file


The signal file is chosen after creation of a new transcription or by changing the signal linked to the transcription with menu [File]/[Open audio file...].

4.2 Sound file format


Most standard audio files are automatically recognized thanks to Snack library: WAV, AU, SND, AIFF, SMP, and Sphere if the format has been installed with Snack. Access to MP3 format can be slow due to the compression. Headerless sound files format (8 or 16 bits sample length, byte order...) is automatically guessed, except for sample frequency and channel count which are given by the user in the window [Options]/[Audio file...].

4.3. Signal shape


In order to speed up signal display at lower resolutions (display of intervals over than 30 seconds), a low-resolution shape of the whole signal is computed if the signal is longer than 30 seconds.

Shape computation is performed in background the first time a signal is displayed, and is stored in a specific directory for subsequent uses. Current computation can be stopped, or the function can be disabled in [Options]/[Audio file...]. Background computation mode can be disabled.

The directory for storing shapes is also given in [Options]/[Audio file...]; it can be a good idea to use a directory shared by all Transcriber's users but it must then be writable to all users in order to allow creation of new shapes.

4.4. Open a remote file


The client-server architecture for sound display and playback can be enabled by configuring the script tcl/SoundServer.tcl and launching it on the host. User choice of remote server and port is done within [Options]/[Audio file...].

5. Signal management


During transcription, one has to listen several times to short portions of signal. Tape-recorder-like commands are available in the command button bar as long as in the menu [Signal], but keyboard shortcuts will probably more practical for most intensive use.

5.1. Position the cursor


One can place the cursor in the signal with a mouse click on the wished position. A precise value can be given with menu [Signal]/[Go to...]/[Position] or by clicking on the information line under the signal.

5.2. Select part of the signal


Part of signal can be selected by click-move-release of the left mouse button, and it can be extended with central button (or shift-click of the left button). Cursor position and selection boundary times are given under the signal. With option Automatic selection playback option on in [Options]/[General...], selected segment playback starts as soon as mouse button is released.

5.3. Signal playback


With play and pause buttons, command [Signal]/[Play/Pause] or its shortcut Tab, one can start playback at cursor position or stop it. If a selection is active, only this part will be played.

The menu [Signal]/[Playback mode] allows a choice between the following playback modes: continuous play (default mode), playback with pauses at segment boundaries, with a beep at boundaries, with a pause at the next boundary, or loop on the current selection if active, otherwise on the current segment after a pause. The duration of the pause and the selection of an audio file for the beep are customizable in the window [Options]/[Audio file...]; the parameter Go back before playing allows to rewind before each new playback.

When Snack version 2.0 is installed with Transcriber, it is possible to play only the left or right channel from a stereo file using the new menu [Signal]/[Stereo channel]. It has no effects for mono sound files or when previous Snack version 1.7 is used.

5.4. Segment playback


Command [Signal]/[Play segment] or its shortcuts (Shift-Tab or Alt-Tab) launches playback of the current segment from its beginning up to its end.

5.5. Fast forward


Fast forward and backward buttons (menu [Signal]/[Go to...]/[Forward] or [Backward]) make the signal cursor move half a second forward or backward. If mouse button is kept down on the command button, move becomes continuous with 0.5 second step. This does even work during playback - but sound becomes choppy.

5.6. Go to next segment


Next and previous buttons (menu [Signal]/[Go to...]/[Previous] or [Next]) move cursor to the beginning of next or previous segment in the orthographic segmentation (even during playback which jumps to its new position).

5.6 Change the sound level


Sound level is given (from 0 to100) in the window [Signal]/[Control panel].

5.7. Vertical zoom


For signal with low dynamic, a cursor in [Signal]/[Control panel] changes vertical amplitude on the signal. Volume is not modified.

6. Edition


6.1. Cut/Copy/Paste


Standard editing commands (menu [Edit]/[Copy], [Cut] or [Paste] and standard shortcuts Control-x/c/v) can be used to handle parts of the transcription, including events but not section, turn or background buttons.

6.2. Undo


Undo command (menu [Edit]/[Undo] or standard shortcut Control-z) is limited, and only restores orthographic transcription of the current segment to its previous state. A lot of actions are not undoable. Save often.

6.3. Find/Replace


Command [Edit]/[Find/Replace] (shortcut Control-f) has various options: case sensitive, backward search, use regular expressions. A replacement string can be given.

6.4. Glossary


Command [Edit]/[Glossary] or its shortcut Control-k has two modes: create a new entry in the glossary, or insert an existing entry in the editor at current position; modification of existing entries is possible in both modes. If part of the text is selected when the command is called, creation of a new entry is proposed from this selection; a comment can be added and choose OK in order to register it. If no text was selected, one can click and select an entry in the list then click Insert; the entry will be inserted in the current position. Glossary is currently kept within configuration file but will probably switch to a separate file in a future version.

6.5. Automatic space handling


Automatic space handling is controlled within [Options]/[General...]. If it is active, a space is added at the end of line upon creation of a new breakpoint and before punctuation symbols .,; or :.

6.6 Spell checking

Spell checking is started by the command [Edit]/[Spell checking]. It relies on the installation of the aspell software (standard on most Unix systems but not with Windows). The dictionary used for the correction is automatically deduced from the language selected in the global options . An error message indicates if the dictionary was not found or if the name does not match; it is then necessary to check your local installation of Aspell to determine the correct name of the dictionary.

When an unknown word is found, a dialog box proposes to:

An option of the menu [Options]/[General] allows to control the spell checking on words starting with a capital letter. The default behavior is to ignore them.

6.7. Accented characters and internationalization

The character input method is not selected in Transcriber but is system-dependent. Thus, with an appropriate keyboard, some characters can be input directly. Unix system allows you to remap some keyboard keys (like for instance function keys) with software tools like 'xmodmap' for X11. However it is also possible to input accentuated characters from keys combinations by defining shortcuts in Transcriber .

The default encoding of transcription files is ISO-8859-1 (Latin1), but other encodings are available if Transcriber is run with a version of Tcl/Tk at least 8.1.1.

7. Episode


7.1. Edit global parameters


Some global informations about transcription can be given in [File]/[Edit episode attributes...]:

8. Segmentation

8.1. Insert a breakpoint


Command [Segmentation]/[Insert breakpoint] or its shortcut Return create a new breakpoint at current cursor position in the text and in the signal. Boundary positions are rounded to the nearest milli-second, so any new breakpoint must be at least 1 ms away from existing ones. A new line appears in the transcription, and segment is sliced at cursor position. This can be done during playback.

=>

8.2. Suppress a breakpoint


Command [Segmentation]/[Delete breakpoint] or its shortcut Shift-Backspace suppress the breakpoint at the beginning of current segment and fold its content with previous segment. If the breakpoint was also the beginning of a new turn or a new section, they are also removed.

8.3. Move the boundary between two segments


By clicking with central mouse button (or with left button with Control key pressed) over a boundary between two segments and moving the mouse, one can move the boundary and the one which are synchronized in other segmentations.

8.4. Select the signal over one or several segments


With a click of the left mouse button over a segment under the signal, the signal is selected over the extent of the segment, and signal cursor moves at the beginning of the segment. The selection can be extended with Shift click of left button.

8.5. Hide/Show a segmentation


With a click of right button, a contextual menu appears and gives commands for signal playback, resolution, and showing or hiding each segmentation.

9. Speech turns and speakers


9.1.Create a new turn


Command [Segmentation]/[Create turn...] or its shortcut Control-t creates a new speech turn from the beginning of current segment. A speech turn is described by it(s) speaker(s) and other optional parameters:

First line of the window is for overlapping speech (see next section). Central part is for creation, selection or modification of speakers. By clicking on Create Speaker or by typing directly a new name, a new speaker is created; one can further describe its type, dialect... A sorted list of already defined speakers appears to the right, with a selection of recently used speakers. By clicking on a name, the speaker is chosen as speaker for the turn. Spelling of its name and other characteristics can be modified after a click on Modify Speaker, and then validating with OK; the changes will apply to the whole transcription. The field no speaker can be chosen for signal segments without speech.

9.2. Overlapping speech


During creation or modification of turn parameters, overlapping speech can be activated. Second speaker can be chosen the same way as the first one. In the text editor, each segment contains 2 lines beginning with 1 or 2; orthographic segmentation under the signal is split horizontally for each speaker. Return to non-overlapping speech is done by deactivating the overlapping speech field in turn parameters; only first speaker is kept and transcriptions of both are merged in one single text.

9.3. Edit or suppress a turn


Command [Segmentation]/[Edit turn attributes...] allows for edition and destruction of the speech turn which contains current segment (even if the current segment is not the first segment of the turn). Alternatively, one can click on the turn button in the editor window. All parameters given during creation can be modified. A button allows destruction of the turn.

9.4. Find one speaker's turns


Command [Edit]/[Speakers]/[Find speaker] displays the list of speakers in the transcription. By clicking on the name, number of occurrences of this speaker in the transcription is shown under the list. Next button moves to the next speech turn of the speaker.

9.5. Import speakers from another file


Command [Edit]/[Speakers]/[Import from file...] asks for a transcription file and displays the list of speakers within this file which doesn't yet exist in the current transcription. User can select/unselect speakers by clicking on its name, or select one speakers which Global name field is on.

9.6. Suppress unused speakers


Command [Edit]/[Speakers]/[Remove unused speakers] suppress any speakers which doesn't appear in any speech turn.

10. Sections and topics

10.1. Create a new section


Command [Segmentation]/[Create section...] or its shortcut Control-e creates a new section at beginning of current segment. Section type ("report", "filler" or "nontrans") and topic can be chosen in the dialog window. Already defined topics list is on the right, recently used ones are on the left. Click New topic or type in directly a text to define a new topic. If an existing topic is selected, one can click Modify topic and edit it, then choose OK. You can choose no topic or empty the Topic field. A new speech turn is automatically created at the beginning of the section; the current speaker is proposed as default value.

10.2. Edit or suppress a section


Command [Segmentation]/[Edit section attributes...] allows edition and suppression of the section which contains the current segment. One can also click on the chosen section button in the text editor. All section parameters can be modified, and a button allows for section destruction.

10.3. Find sections about a topic


Command [Edit]/[Topics]/[Find topic] displays a list of already defined topics. Click on a topic and its count will appear under the list. Next button moves to the next section about the topic.

10.4. Import topics from another file


Command [Edit]/[Topics]/[Import from file...] ask for a transcription file and displays a list of new topics. User can select/deselect topics by clicking on their name.

10.5. Suppress unused topics


Command [Edit]/[Topics]/[Remove unused topics] suppress unused topics from the list.

11. Background noise

The segmentation in acoustic conditions if for lasting modifications in background noises. The marks are for a change in background state; their value replace the previous one. Each transcriptions begins with an empty background condition at t=0.

11.1 Create a background mark


Command [Segmentation]/[Insert background] creates a background mark at current cursor position in the signal. It is not allowed to create two marks at the same time index. If the cursor in the signal is at the beginning of the segment, background mark will always be synchronized with segment beginning, and background icon appears at the beginning of line. Background type (speech, music, electric noise, other) is chosen; value defaults to previous one. A music icon appears in the editor at current position of the cursor. Background segmentation under the signal is colored (grey default color) on signal parts with noise.

11.2. Indicate the change or the end of a background noise


Create at the position of the end of noise a new background mark with all noise types empty. For a noise change, don't mark the end of the previous noise, but only the new one which replaces the previous one.

11.3. Edit or suppress a background noise mark


Background noise marks can be edited or suppressed by clicking on the music icon in the editor. If it is deleted, current background conditions become the ones defined in the previous background mark.

12. Events and comments

12.1. Insert an event


Command [Edit]/[Insert event] or Control-d pops up a window for defining:

12.2. Insert the beginning and the end of an event


If a part of text is selected before the insertion of the event, beginning and end of event marks are simultaneously inserted

12.3. Insert a comment


[Edit]/[Insert comment] inserts a comment event. Comments have no temporal extent.

12.4. Edit or suppress an event


Click on the event in the text editor and edit or destroy the event. Events can also be deleted in the text editor like a single character.

12.5. Copy/Paste an event


Within Transcriber, events can be copied/pasted. When pasted in other applications, event marks will appear as XML tags.

12.6. Create a keyboard shortcut for an event


See shortcut section

12.7. Edit the default event list


Noise list, pronunciation marks, lexical marks or language list can be edited with [Options]/[Events]/[Edit * list...]. Value will appear in the editor and in the segmentation; description appears in the menu with list of predefined events. An empty line is a separator in the menu. For language list, iso639 codes should be used for consistency.

12.8. Configure events display


Fonts and colors for event display can be modified with menus [Options]/[Fonts]/[Events] and [Options]/[Colors...] ; format string for comment and events, and for event extent are chosen in window [Options]/[Events]/[Events display...].

13. Keyboard shortcuts

Menu [Options]/[Bindings...] pops-up a window for edition, suppression or creation of bindings.

13.1. Define a shortcut


Click New in [Options]/[Bindings...] for creation of a new shortcut, then type in keyboard combination in the first field, replacement string in the second and click OK (end of edition) or Modify (before other editions). Keyboard combination can be directly typed with Control, Shift, Alt... if cursor is in the Keystrokes field. Be careful, you can override predefined menu shortcuts.

13.2. Shortcut for a currently used expression


Select currently used expression with mouse in the text editor, then choose menu [Options]/[Bindings...], give wished Keystrokes and click OK. Shortcut can also consist in the first letters of the expression, e.g. "gor" will be replaced with "Gorbatchev". These shortcuts are not related to the glossary.

13.3. Shortcut for an event


Events (noise, comment, etc.) can be generated with a shortcut. Create an event of the wished kind in the text editor, select it and follow the previous instructions. Selected text can be any combination of regular text and one or several events.

14. Options and interface configuration

14.1. Transcriber's name


Default transcriber's name is given in [Options]/[General...]. It is used in episode parameters initialization when creating a new transcription.

14.2. Productivity monitoring


A filename for transcriber's productivity monitoring can be given in [Options]/[General...]. By default, the field is empty and function is disabled. When it is active, work duration is traced (without pauses), as long as produced work (number of new sections, turns, words...); it is displayed in window [File]/[Informations] and saved in the logging file at the end of the session. No automatic exploitation of these file is currently performed by the software.

14.3. Colors edition


Most colors can be modified with menu [Options]/[Colors...].

14.4. Fonts edition


Fonts used for text display are modified with menu [Options]/[Fonts].

14.5. Localization of the interface


Language for the interface is chosen in [Options]/[General...]. It defaults to English, and French is available. Localization in new language can be created. When no localization is available, the original English message is kept. As soon as non-localized messages are displayed, they are added to the list of messages to be localized. In order to keep your localization, give a Localization file name in [Options]/[General...] and choose [Options]/[Save configuration].

14.6. Second signal view


Menu [Options]/[Display]/[Second signal view] shows or hide a second view of the signal, which can be displayed at another resolution. Resizing the window can be necessary for a correct display.

14.7. Show/Hide central button bar


Central button bar display is switched on/off with [Options]/[Display]/[Command buttons].

14.8. Save the options


Modified options, window positions, keyboard shortcuts, glossary content and currently edited filenames are saved with command [Options]/[Save configuration] in a configuration file (defaults to ~/.transcriber); this file is read at startup and currently edited files will be automatically reopen at beginning of next session.

15. Other features

15.1. Informations about the signal and the transcription


Button or command [File]/[Informations] displays a window with informations about signal filename, duration, word count, etc. Click update button for an update of the informations.

15.2. Quit Transcriber


Menu [File]/[Quit] or Control-q ; save of the transcription is proposed if necessary. If corresponding option is activated, save of the configuration is also proposed.

15.3. Help


Online help is available with menu [Help] ; help files can also be viewed with any HTML browser with frames