Interviewing First Gen. Scholars about the academic year of Covid (2020-21); some reflections on data collection using Semi-structured Interviews conducted online
Blog by: Dr. Lewis Mates, School of Government and International Affairs (SGIA), Durham University<br>
N.B Part #4 (and #2) is on the PSA website; Part #5 will appear here in June 2023.
Auto transcribing has improved… but it still isn’t perfect….
One of the attractions of Teams was the auto-transcription function, that I used to try to avoid the costs of transcription services (I had only a small research budget).
The early vtt texts produced this way were, however, awful; three or four words of transcription and then a couple of lines of gibberish (totalling maybe 60% or more of all the entire transcript) like this:
- NOTE Confidence: 0.72214794375
- 01c24cfc-1ecb-4f28-a452-abcdd49ae4e1
- 00:00:00.000 –> 00:00:01.304
Because it’s suddenly give
NOTE Confidence:
- 0.72214794375
da3d2dc3-dd40-4ff1-aad8-69c94385cb08 - 00:00:01.304 –> 00:00:03.082
me a different option. Oh,
When confronted with a mess like the above, I tried to clean them up manually but that was a time-consuming and soul-destroying process, not least as the sentences that finally began to emerge often didn’t make much sense either.
I also found that the longer interviews would not auto-transcribe at all. I therefore had to halt the recording of interviews about an hour in, and then restart the recording, in order to give the auto transcription the best chance of working. Those that I hadn’t split this way, weren’t initially transcribed at all.
Fortunately, Teams improved rapidly and the later interviews for the same project got auto-transcriptions rather more like this (and without the need to stop recording after about an hour):
- 00:40:24.520 –> 00:40:26.340
LM: Impressive degree, don’t you so?
- 00:40:25.950 –> 00:40:29.740
‘Amy’: I mean, we did do it to ourselves. We did choose this degree, so shouldn’t shrink complain.
(Here I have just anonymised the student’s name). While this is clearly much better, the transcript is obviously still not perfect and needs a bit of work to interpret and check etc.
The early Teams recordings I ran through Panopto recently and got a transcript similar to the second example in quality (so much cleaner and more accurate) but obviously with no sense of who is speaking. Going back through them, however, it was easy enough to work out the speaker and to delete lots of extra conversational words that were of in use for the project. Where necessary I précised what I had said in square brackets, to offer the analyst a sense of the context in which an interviewee had said something interesting.
Naturally, the money and time saved by not using a professional transcriber is not to be sniffed at. And some of the auto-transcriptions can be pretty amusing;
But basically my family, my dad sounds pretty much fully Jodie Foster sound. Really, Jodie. They’ve lived here all their lives. [‘Jodie’ is Geordie; no idea where the ‘Foster’ came from…]
and
and
and
and finally…
In spite of power dynamics, interviewees will volunteer critical as well as positive feedback of you, the interviewer…
When discussing their experiences of teaching and learning during the Covid year, some first gens would comment directly on classes they had taken with me, or on the work I had done in trying to support first gens more widely.
As a lecturer and often the actual teacher of many of the students I was interviewing (albeit after they had finished their year with me and exams were done etc.,) the power dynamic was still clearly rather different from, say, a PhD student conducting ‘elite’ interviews for the first time. Acutely conscious of this, I tried to say to all interviewees at the outset that they needed to be as truthful as they could be and if that involved pointing to my own deficiencies that year, then they had to do so, and not hold back.
Quite a few interviewees were fairly positive about my own teaching etc., and I did wonder whether they might have said something different to another interviewer (though my student module evaluation scores that year were fairly good as well, I suppose..).
But there were criticisms too; the strongest was from an interviewee who quite rightly objected to an incident (that I’d forgotten) when I’d asked about her possible first gen. status because I had noticed she had a slight regional accent. The interview gave me the chance to apologise there and then and resolve to be more sensitive in how I framed such conversations in future. Analysing that particular interview recently, I realised again how especially rich and insightful it is.
9. Try to not do too many interviews in a single day if you can…
Due to time constraints I sometimes had to do several interviews a day, though this can cause problems. I found myself telling ‘Barbara’, for example;
LM: This is the third of these interviews I’ve done today. So if I say something and you think “hang on a second, I didn’t say that”, could you please, like, alert me to it? Because what it would be is I’ve kind of messed up a memory of something that someone else said earlier in the day and not you. So, yeah, […] it’s suboptimal, but, you know, we’ve got this small period of time once exams finish before summer holidays start where we can do a bit of research. So it’s like action stations, you know, anything you want to do research wise, now’s the time to do it. So if you don’t kind of like make the most of it, then yeah, you’re going to be kicking yourself the rest of the year thinking, “Gah, I wish I had a bit more data now I could work on…”
Blog by Dr. Lewis Mates, School of Government and International
Affairs (SGIA), Durham University lewis.mates@durham.ac.uk
Contact
Questions about membership, publications, events and other FACE activities should be directed to the team at webmaster@face.ac.uk
Monthly e-bulletin
Both members and non-members can sign up for our free monthly e-newsletter by emailing webmaster@face.ac.uk