Transcription is a crucial part of conducting research
In this case study, you’re going to read how transcription time dropped from 3 weeks to 5-7 working days. Secondly – you’ll know how accurate manual transcription services are. Certainly, this is a crucial part when conducting, processing, and publishing research.
The way transcribing used to be done
Amsterdam’s University of Applied Sciences (HVA) research team CAREM conducts many kinds of research. In most cases, it involves recorded interviews. After that, they had to be transcribed in order to be used in research.
Before involving Amberscript into the process, it looked like this:
- The team conducted & recorded an interview. This results in a significant number of hours of recorded audio.
- At a certain point in time, usually, when the research conducting phase was coming to an end, an email notification was sent to (work) students. They were offered to manually transcribe the recorded interviews for an hourly rate.
- This meant an opening of the hiring process for all students that were interested in helping to transcribe audio to text.
- As a result – all invited students had to be informed on how to transcribe.
- After that researchers had to conduct a quick check of transcribed texts, as we need to keep in mind that students aren’t seasoned transcribers, and their work sometimes requires feedback.
- Transcriptions were sent back to students for adjustments and improvements.
- Finally, finished transcriptions were delivered and could be used in the research.
Sounds like a time-consuming process? In fact, it was – from start to finish all took approximately 3 weeks.
Downsides of manual transcription process?
- The process takes a long time before transcripts are ready. As a result, this slows down the research as a whole.
- Involving researchers into recruitment, hiring, and training, not only takes away their focus from the research. It also costs them a lot of time and energy.
- Students are not always reliable transcribers. During the process, they can quit at any time which endangers research deadlines even more.
On a positive note: students transcribe audio to text at a relatively low hourly rate.
Amberscript’s automated transcription service
Given all of the above, Amsterdam University of Applied Sciences decided to start using Amberscript’s transcription services to convert recorded interviews to transcripts. Results:
- The time needed to transcribe decreased – from 3 weeks to 5-7 working days.
- The hourly rate of transcription remained more-less the same. What’s most important is saving time compared to the old transcription process.
- The process itself is much more simplified than it used to be. No hassle with recruiting students, training, and giving them feedback.
- Accuracy remained good in case of a mistake, Amberscript is always flexible to correct them in a timely manner and without any additional costs.
- The new workflow allows researchers to focus on the tasks that really matter to them.
Selecting Amberscript as a partner in transcribing speech to text
When looking for companies that offer transcription, these services seemed to be out of budget. Amberscript was able to give a competitive offer, which enabled Amsterdam University of Applied Sciences (HVA) to outsource the transcription process and build a sustainable partnership.
Amberscript even offered to hire some of the students that worked for the HVA, that way they did not have to look for a different job next to their studies.
A new way of converting recorded audio to text
The whole transcription process has been simplified. Therefore, it requires much less attention than it used to:
- A team of researchers conducts & records an interview.
- Recorded data is handled over Ria van der Holst – the main contact person between HVA and Amberscript.
- A source file is uploaded to our online editor.
- Amberscript’s AI engine does the first transcription (it’s usually 85%-95% accurate). This results in a clean-read transcription
- Each source file goes for a manual check conducted by a transcriber, to make it perfect (99% – 100% accuracy).
- Source files are uploaded to an online editor for a final check.
- Outcome transcriptions are available for download. It can be exported to any supported text file format (in most cases it’s .docx, file opened with Microsoft Word editor).
To sum it up
Analyzing a large volume of voice recordings, without transcription would be an impossible task to handle. Thanks to using the transcription service everything is possible. We’re very happy to become a partner of Amsterdam University of Applied Sciences, in handling this important task.