Transcribing user interviews – speeding things up
We’ve been conducting lots of user research interviews recently. We transcribe our interview recordings to more accurately reflect what people said but this comes with an overhead. We’ve been trying automated transcription tools to reduce the effort and speed up our work.
User research is fundamental if we are to make sure that we create content that actually answers the needs of the prospective students and the staff who support them. We’ve been talking directly to prospective students (as we did in our guerilla research at the Postgraduate Open Day), and more recently we have been interviewing colleagues whose jobs involve regularly dealing with their enquiries.
Understanding users is our number one principle – read more about principles
Conducting guerrilla usability testing at Postgraduate Open Day
Why do we transcribe our interviews?
When we interview someone and only take notes, a few things happen:
- We’re not giving the interviewee full attention as we’re also trying to write
- We capture only a fraction of what has been said and done as we can’t keep up
- What we write down goes through the filter of our own brains, so we introduce our own biases by choosing in the moment what to record and how to record it
As well as avoiding these issues, recording and transcribing interviews also opens up opportunities for collaboration and better shared understanding across our team. Our User Experience Specialist, Gayle, has been working with us to draw out key comments and then organise them to identify trends across multiple interviews.
Having the full transcript of the interview has meant that I could rapidly scan everything that was said in someone else’s interview, without having to listen to the full interview recording. If we didn’t have a transcript and were just relying on notes jotted down during the interview, we’d risk either entirely missing out useful statements from our notes, or even if we’d noted a useful comment, we might not have captured all the detail and nuance contained in the exact words used by the interviewee.
This gives us much more robust insight which helps us improve our content planning and ensure it’s more evidence based.
Automating transcription to speed things up
I’ve only recently joined the team, so haven’t taken part in any of the interviewing yet, but have been involved with some transcription and subsequent sensemaking sessions (these are where we take all the relevant comments from the interviews and synthesise them, grouping similar issues together).
As we’ve been working our way through writing up the interviews, we’ve each experimented with various transcription tools and compared results.
Out of the ones we tried, we found a service called Temi best suited our needs.
We’re always interested to find better tools though, so if you’re reading this and use a different option for transcription, comment below with a recommendation. We’d love to hear about your experiences.
What is Temi?
Temi is a transcription app, using advanced speech recognition software to provide speech to text transcription.
Why we like it
Here’s why we like it so much:
- They accept all audio or video file types.
- They have a fast turnaround time. They promise on their website to send your transcription in five minutes. In our experience so far they’ve made good on this commitment.
- When you receive your transcript it’s already marked up with timestamps and speakers. However, we’ve noticed the software sometimes struggles if both speakers have a similar tone to their voices.
- The audio is embedded with the text, so it’s easy to listen again to a specific phrase that you think the software might not have recognised correctly. You can adjust the playback speed and skip around the text without ever getting confused about where you are in the recording.
- You get an initial free trial of an up to 45 minute audio or video file transcription. After that, it’s $0.10 per audio minute, with no subscriptions, no minimum amount of minutes and no hidden extra charges.
- You can download your transcript into text (MS Word or PDF) or closed caption files (SRT, VTT).
What if your audio isn’t that clear?
The Temi website says they need clear audio with little background noise, clear speakers and minimal accents. However, they actually manage surprisingly well with much trickier recordings.
An interview that I recently transcribed was recorded in ECA café. There was lots of background noise and the two speakers regularly talked over one another. Both people had reasonably strong regional accents. The software got confused sometimes when the speakers spoke across one another. But overall the transcript was a remarkably accurate record of the spoken conversation.
Does it do everything for you?
As with all such transcription services, after you receive the transcription, there’s still a job to do. You need to edit out all the ums and ahs. Sometimes the speakers repeat themselves or someone sneezes in the background and drowns out their voice. All that aside, the service takes out the grind of inputting the entire interview in the first place.
Tell us about your experiences
We’re always interested to find better tools though, so if you’re reading this and use a different option for transcription, comment below with a recommendation. We’d love to hear about your experiences.
Thank you so much for this blog post. I just had a test run through with Temi and it’s amazing, it’s going to save me so much time transcribing interviews. Thank you.
nice article!
Although the usefulness of video material cannot be overstated, transforming audio into text transcripts with Transcriberry may provide many more benefits in terms of marketing and teaching.