Department of Cybernetics @ University of West Bohemia in Pilsen

Structural Metadata Annotation for Czech

QAn - MDE Annotation Tool

To ease the MDE annotation process we have developed a new specialized software tool respecting Czech particularities and the modifications of rules we made - QAn (Quick Annotator). Similar to the English MDE tool, it enables to highlight relevant spans of text, play corresponding audio segments, and then record annotation decisions with few mouse clicks or keystrokes. The program can be [DOWNLOADED HERE].

Figure 1: QAn screenshot

Installation and file formats

The program can be runned under Linux OS. To run it on your computer, you need to have the following packages installed: Installation is very easy. Just extract the downloaded archive to your filesystem and run the script "" from the QAn's main directory. After the first start, the program will create the directory ".qan" in your home directory and copy there the ".ini" files from the main directory. The program enables to open a transcription file in the Transcriber format (.trs). The metadata annotation is saved in a format that enriches the baseline .trs format by special MDE tags. This format is called ".qan".

QAn displaying style:

On the screen, the transcript is displayed divided into turns separated by solid lines. Turns are numbered and labeled by speaker names. Similarly, each turn is divided into segments according to the breakpoints existing in the original transcript. Each such segment starts on a new line. Note that the regions of overlapping speech are displayed in grey and cannot be annotated or edited; they are not used for structural metadata annotation.

Regular tokens

words  -  are displayed using the initial font
wordfragments-  -  are denoted by -
*slips*  -  are displayed in pale green within asterisks
[NON-SPEECH EVENTS]  -  are displayed in gray italics within square brackets


(boundaries of Syntactic/Semantic/Sentential Units) are displayed in red:
/.  -  Statement break without strong prosodic marking at boundary
//.  -  Statement break with strong prosodic marking at boundary
/?  -  Question break without strong prosodic marking at boundary
//?  -  Question break with strong prosodic marking at boundary
/-  -  Incomplete (arbitrary abandoned) SU
/~  -  Incomplete SU interrupted by another speaker
/,  -  Clausal break - identifies non-sentence clauses joined by subordination
/&  -  Coordination break - identifies coordination either of two dependent clauses, or of two main clauses that cannot stand alone


Discourse markers  -  (DMs) are displayed in dark brown
Discourse responses  -  (DRs - a subtype of DM) are displayed in light brown
Explicit editing terms  -  (EETs) are displayed in cyan
Filled pauses  -  (FPs) are displayed in dark red (if annotated using FP labels rather than as non-speech events)
{Asides/parentheticals}  -  (A/Ps) are displayed in gray underlined italics within curly braces

Edit disfluencies

[Delreg]  -  Deleteable regions (DelRegs) are displayed in italic green within square braces
[Delreg * Delreg]*  -  Interruption points (IPs) are displayed as red asterisks
Corrections  -  Corrections are displayed in bold black underlined italic

QAn control:

Controlling QAn is easy. First, use the 'Set wav dir' from 'File' menu to select the directory in which your waveform files are stored. The waveforms must always have the same basename as the transcript. The transcript (in .trs or .qan format) can be opened from the File Menu. To play audio for the corresponding segment, just select any word within the desired segment and hit the spacebar. The playback can either be stopped by hitting the spacebar again, or you can move in the audio using arrow keys on your keyboard. Everytime you hit - or -, the playback automatically moves by 3 sec in the corresponding direction. Alternatively, you may use the slide bar located below the displayed text to move in the audio.

To mark fillers and edits the annotator simply selects the desired span of text using his/her mouse and then assigns to it the appropriate label either using a keyboard shortcut or selecting from the context menu.

Figure 2: Annotation of labels (fillers and edits)

Annotation of SUs and IPs is very similar to annotation of fillers and edits. The appropriate symbol is inserted right after the last selected word. Alternatively, the annotator may select a punctuation symbol and replace it by an SU symbol easily.

The tool also enables to words in the transcripts. If the annotator finds an error (missing word or non-speech event, typo, wrong spelling, etc.), he/she should correct it. Insertion of non-speech events (most often filled pauses) is enabled in the context menu. To replaceme incorrectly transcribed word(s), one must select the word(s) of interest and then replace/delete it(them) by choosing the item "Substitute words" from the context menu. The words may be edited in the popped-up dialog box.

Figure 3: Transcript correction

QAn is fully configurable and can be customized easily by editing 2 configuration files. The first enables to customize shortcuts, the second enables to add/remove types of annotation symbols and to change their displaying style (colors, automatic insertion of IPs, etc.).

An example of a structural metadata annotated transcript follows:

Figure 3: Larger example of an MDE transcript

You can download a sample transcript along with a corresponding audio file.

The program can be downloaded [HERE].

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see