HCI research: 9 tips for doing transcriptions
“ I don’t have time for this”, “This is going to be so boring! I don’t want to do it!” and “Aren’t there any apps for this?” These are some of the thoughts that crossed my mind when I was told that I was going to have to do transcriptions for a study.
Transcribing is often seen as a long, time-consuming, and tiresome task that no one wants to do. Fortunately, someone wise told me (paraphrasing): “You should do it yourself. You will find that, as you do it, you will get acquainted with the data, you will start seeing patterns, and that will improve/enrich the whole work”. Still, I wasn’t convinced, and the thought of spending days on end doing the same, monotonous task was freaking me out. Yet, I had no choice. I could not afford to hire a service or buying an app, and I had no one who I could delegate the task to. So I did it myself, and I am glad I did. In the end, it did help me to get more intimate with the data and ended up facilitating the ensuing steps. I used to see transcription as a boring, isolated task you do before analysis, but now I understand that transcription is an actual component of the analysis itself. Furthermore, I think that transcriptions are an invaluable part of the process of better understanding the users and their needs.
Get familiar with the whole process before starting
Common Human-Computer Interaction (HCI) methods like user interviews, participatory design sessions, or focus groups, are usually recorded (audio or video) so that the data can be analyzed later through qualitative data analysis methods, and a frequently used one is Thematic Analysis. That being said, it helps to have a basic understanding of what a Thematic Analysis is before doing transcriptions.
Basically, a Thematic Analysis is a method that helps you organize all the research data you collected previously and make sense of it. In a Thematic Analysis, you will be constantly moving back and forth the entire data set to find and code patterns and themes, that will later help you define design implications. By using this systematic approach properly, you will be able to ensure that your results are valid.
The first step in a Thematic Analysis is to get acquainted with the collected data, and there is no better way to do that than by doing the transcription yourself if you can. As you listen to the recording you will begin to make sense of the data and naturally start seeing patterns in the information, saving you time in the future and helping you to be more intimate with the data.
Be diligent when recording the data
If the quality of the recording is poor, it will take you much longer to do the transcriptions, so make sure that the recording is well done. You will want to have a sharp audio recording, where you can clearly listen to the people talking, and the best way to guarantee that is to test the recording in advance in the same room where the study sessions will take place (or similar) — you can, for instance, ask a colleague (or just do it yourself) to sit on the end of a table and talk for a bit while you record it. Then listen to it and see if the sound quality is good. Also, check for possible sources of noise (open windows, a projector whirring, etc.)
Block time after the study sessions to do the transcriptions
Transcriptions can take a really long time (“at least 3 hours per hour of talk and up to 10 hours per hour with a fine level of detail” [1]). This can mean several days (or weeks) of work. The lack of experience of the transcriber, the quality of the audio recording (video recordings take longer since you have to analyze facial expressions, etc.), background noise, and whether is one person or multiple people talking, can make the period even longer.
Consider 7 or 8 hours for each hour of recording to estimate the time you will need to do the full transcription. So, if for instance, you have 10 hours of audio, you should block 80 hours plus time for breaks on your calendar for this task. Keep in mind that we usually overestimate how much focused time we have in a typical day, and this is a task that requires much focus — so go ahead and block extra time!
Plus, you should do the transcriptions as soon as possible because…
Timing is everything!
Transcribing should be done soon after the study sessions, while you still remember the participants and their voices — it will be easier to analyze and transcribe if you still remember who said what and why. You will still remember details that you might forget later.
Furthermore, if you can make sense of some part of the recording, you can contact the study participants at this time and clarify with them what they said (after a while, it is unlikely that they will remember what was said and why).
Decide on a notation system
Transcribing involves a lot of decision-making. If you don’t prepare ahead, you will most likely be flooded with questions like:
- Should I write fillers like “hmmm” or “oh”? What about stutters?
- Should I consider transcribing sounds like laughter?
- Should I take note of pauses? Should I write how long the silence lasted?
- How should I indicate overlapping conversations (two or more people speaking at the same time)?
- How do you mark unintelligible speech?
To avoid being constantly interrupted by questions like these, it is important to choose a notation before starting.
When choosing a notation system you should try to find a balance between the time it will take you to write with precise detail (True verbatim) and the value that those details can provide. Keep in mind that, in the end, what you want to achieve is a better understanding of the participants and to produce insights that will guide the design.
One example is the Jeffersonian Transcription Notation [2].
Whether you use an existing notation system or you create your own, make sure that you agree upon it your team beforehand.
Also, it is good to complement the transcription with comments about your interpretation of tone (e.g., “I think she was being ironic when she said this”) or to note useful impressions you remember from the study sessions (e.g., “I remember that he was doing a lot of hand gestures at this time”).
Do your best to make the transcripts easy to read
You and your team will look at the transcripts all the time — you will read them over and over again to reflect, find patterns, extract quotes to support your findings, and so on. Thus, transcripts should be easy to read and navigate.
You can start by adding a header to the document with relevant information, such as:
- Name of the project
- Information about the research team
- Date, time and location of the recording of the event
- Participants, identified by pseudonyms or codes
- Name of the audio file
- Link to the notation system used
- Notes
Then, as you write, make sure that you keep the document clean and structured — choose appropriate values for fonts, margins, spacing, paragraphs, etc., so as to make it as visually efficient as possible. Everyone (including yourself) will thank you later.
Tools of the trade
There are several tools that you can use to help you in the transcription process. Some of the most well-known are:
However, you can do transcriptions using a simple player and a word processor. Personally, I like to use the VLC player and Google Docs.
This is my setup (image above): I place the player and the document side by side on the screen and use the “command-tab” keyboard shortcut on Mac to quickly toggle between both apps. VLC allows you to use shortcut keys to pause and move backward through the audio (step backward). It also allows you to add bookmarks, in case you want to mark relevant moments in the recording that you might want to listen to again later or show to someone. You can also try using different playback speeds — this can be really useful if, for instance, one of the participants talks very fast, or when there is an overlapping conversation that you need to unveil.
Try Touch Typing
Touch typing is the ability to type without looking at the keyboard while you do it and using all the available fingers. Touch typing allows you to type much faster, so it is a great skill to learn if you want to do transcriptions more efficiently.
Try it: https://www.typingclub.com/
and lastly…
Do not overthink everything
Qualitative data can provide a deep and extensive understanding of users' wants and needs. Unfortunately, this type of data can be difficult to process, organize and analyze, and generating valid conclusions from it can be troublesome. You must be as rigorous as possible when transcribing, but with all the decision-making involved, if you try to make the perfect decision every time you will probably get very frustrated very fast.
Oftentimes you will have to rely on your instincts and best-judgment. Try to keep the end goal in mind and everything should be fine. In the midst of so much data, it is very likely that you will still extract the right information.
Another tip is to add comments throughout the document to support your decisions when you have doubts. This will make it easier for you and others to understand the data later on.
Sources:
[1] Julia Bailey, First steps in qualitative data analysis: transcribing, Family Practice, Volume 25, Issue 2, April 2008, Pages 127–131, https://doi.org/10.1093/fampra/cmn003
[2] Jeffersonian Transcription Notation is described in G. Jefferson, “Transcription Notation,” in J. Atkinson and J. Heritage (eds), Structures of Social Interaction, New York: Cambridge University Press, 1984.
