CapsQual: Tool for converting ASR output into qualitative interview transcripts

[anouarg88]

Hi there!

I have made a free open source GUI-tool for reformatting output of automatic speech recognition software (such as Whisper) into transcripts following established conventions such as GAT2, Kuckartz/Dresing & Pehl or TiQ. The exact output format is highly customizable. I spent a lot of time on this and would love if someone else also found this useful/would like collaborating. 🌞

It is called CapsQual and can be found here. Please let me know what you think and feel free to contact me. 🌹

Screenshot:

Note: Integration with OpenQDA works best if the content of the exported file is copy-pasted into the document.

Best wishes
Anouâr

[kadewe]

Hey Anouár,
that's a great software!
I really like it and will check it out in more detail, as we have no such functionality in OpenQDA and are also thinking hard about integrating timestamps and linking it to the audio, so that one can listen to the original directly from the transcript.
As we are also exploring using a python subsystem for data analysis i could imagine integrating your tool into openqda. right now we are thinking about ways to do this safely.
thanks for reaching out!
yours
Karsten

[anouarg88]

Hi Karsten,

Thank you very much for your kind feedback! 🌞 I would be thrilled if CapsQual could become useful to you and/or the OpenQDA community! Let me know if there is anything I can do to help.

Best wishes
Anouâr

[kadewe]

Hi Anouâr,
in your documentation you mentioned that import into OpenQDA works best with copy-pasting. What is missing in the import of the exported document? Can you provide us with an example document, so that we can check this?
yours
Karsten

[anouarg88]

Hi again Karsten,

Basically, there's an issue with disappearing spaces that is relevant for GAT2 and TiQ because spaces are used to keep speaker turns lined up (when an equidistant font like Courier New is used).

Here are some example output documents:

Osterhasi_GAT2.docx
Osterhasi_GAT2.html
Osterhasi_GAT2.txt
Osterhasi_TiQ.docx
Osterhasi_TiQ.html
Osterhasi_TiQ.txt

Importing the TiQ docx, for instance, results in unaligned speaker segments, while copy-pasting the content from word keeps the correct format (and the rich text formatting, too, surprisingly). Note that the .html files don't contain newline tags (which is something I should fix in the next update) and so they are not really suitable for import to begin with.

Hope this helps. Should I repost this in issues?

Best wishes
Anouâr

[anouarg88]

Hi! Quick update: in the latest release, I made sure the HTML export works smoothly with OpenQDA import (by using more HTML and less CSS for formatting). It now imports nicely in OpenQDA, except for the font which needs to be changed to Courier New or another monospaced font manually.

Example HTML file:
Osterhasi10.html

In OpenQDA (after changing font):

I also moved the file export and text generation methods to separate .py files for more clarity.
Best wishes
Anouâr

[kadewe]

well, it looks very good and clean now, too bad it needs all the &nbsp - elements.
I wonder if this is a pandoc issue and could be solved by a simple flag… @jankapunkt just to keep you in the loop, i will test this with a local pandoc installation…