Free accurate transcription using MS Word and a cleanup tool
For qualitative research, transcribing interviews or other spoken audio is a pain. Many paid options exist, but it can be a hassle to get approval because academic decision making is painstakingly slow.
In this article
- How to auto-transcribe with Office 365 online
- How to clean up the output ready for analysis
After this, you can easily start coding in Atlas (or other tools). This saved me many hours.
1. Microsoft Word has free transcription in your browser
A colleague of mine tipped me this. Most universities have Office 365, and your online login has 300 free minutes of transcription.
In your ribbon (the toolbar), there is a dictate button, but hidden underneat there is transcribe function.
Note: You should open Word in your browser. My Mac version doesn’t show the transcribe button. If you are not seeing the transcribe button, it could be that your browser is not supported. I used Firefox.
2. First select language, then upload file
The UI isn’t the best. The moment you upload a file, it will start transcribing.
So you better first select the language. It has a lot of languages, English, Dutch, Danish, German, French and more. Check if your language is there.
My experience with quality
I’ve transcribed using English (US), English (UK) and Dutch so far. My recordings are Zoom recordings. It works quite well. I would say there is a hitrate of at least 95%, often higher.
For me, it doesn’t need to be perfect, I will listen to the recording with the text next to me, it’s about highlighting the interesting bits and correcting any typos in Atlas.TI Web.
When publishing certain excerpts, I will relisten and type manually. But that will be less than 1% of everything. The output is very readable by a human.
3. Renaming Speaker 1, Speaker 2 etc
After some waiting, it’s done. You can’t close your browser. But, after transcription is done, it’s linked to that document. So when you see a list of speakers, it’s safe to close your browser.
You’ll see your transcribed text. Perfect? No. But it’s a great first step. The nice thing about this transcription is that you can assign speakers names and change them all at once.
The software is a bit slow, so be patient here. If you click the blue timestamp, it will start playing the audio file from that point.
Tip: Let all participants say their name at the start of the interview/session. This will make assigning speakers much easier
4. Output your transcription as you wish
You can choose how the text is outputted. I always use speakers and timestamps. Helps me to go back to interesting bits.
Sometimes it can take a moment for all the text to appear in your document after hitting the add to document button. Be patient.
However, you can see that MS Word adds a lot of lines. That makes it a lengthy, lengthy document. An hour of interview soon turns into 50 pages.
This also hinders transcription. However, my dear friend who is a good programmer made a simple tool that fixes this. In return he got a bottle of Whiskey.
5. Download the word doc
We now need to get everything on our clipboard. For some reason, downloading a copy is the easiest way. Plus, an offline backup can never hurt.
6. Open the word doc on your computer, select everything and copy to your clipboard
My browser doesn’t let me select everything in MS Word online. So, this extra step is required for me.
7. Clean up with the cleanup tool
My friend made a simple tool to clean it up. Click the link above. Paste your Word transcription and hit ‘Converteer’
This tool doesn’t collect data, it runs in your browser, no information is sent to my server or anything.
You could download the html and run it from your desktop while not connected to the internet if you want to be sure. But it’s client side. Trust me.
8. Paste the result in your coding software
The output in the bottom text field is your tightly formatted text. I copy that right into my coding software.
This is what I get when I paste the output in Atlas.ti Web.
And done. I’ve been able to do this process within 5 to 10 minutes per hour. It makes coding much easier.
That’s it
Thanks for reading, I hope this helps. Let me know if you have any questions.
