For many students, PhD students and researchers, writing a thesis or dissertation will necessarily require transcribing interviews. This article will reveal some powerful tips so that you can effectively transcribe interviews for your thesis or dissertation.
Record your interview
This is meat-and-potatoes for your transcript.
Without audio, there can be no transcript, only vague memories to sum up (unless you are an expert stenographer and have taken notes using the related method… but if that is the case, you are probably no longer a student!). To produce a quality recording of your interview, just, follow these seven steps.
You will recall that the 4th step is to inform your interviewees that they are being recorded. This is a matter of courtesy, but also necessary when conducting so-called “sensitive ” interviews, for instance if you are dealing with patients for research in a hospital setting, etc.
Talk about the recording and give reassurance
In some cases, you will have to convince your interviewees that recording is necessary and reassure them that you will take great care in using, storing and, if required, destroying the audio after use. With sensitive interviews, you will need to make your recording anonymous, by turning all references to the individual’s name into “a patient” for example.
Agree on the working procedure before you begin recording — this can smooth out many a bump in the road.
You will need a quality recording
Good quality audio is “5 by 5”
In military jargon, it is said that good audio is like a message received 5 by 5. Contrary to popular belief, this does not mean, as it did in school, that the audio has achieved a maximum mark of 5.
The two digits in the 5 by 5 represent a scale of values:
- The first refers to the power of the audio signal.
- The second denotes the clarity of the communication.
The closer you are to a signal received “5 by 5”, the better you will ensure that:
- your audio has good volume, and thus that the voices are clearly heard.
Your audio signal will be strong if you have ensured that the recorder, for instance a dictaphone, is placed close to the speaker or if you use accessories such as a microphone to improve recording.
- the voices are intelligible, i.e., you can clearly make out what was said.
This is important: no one wants to end up with frequent “inaudible” passages — the kind which even seasoned veterans cannot make out!
You can guarantee good sound clarity by minimising background noise and/or setting rules for dialogue, so that no voices overlap.
What exactly is transcription?
There are several types of transcription:
- sociological transcription, used for interviews in social sciences, where the idea is to transcribe exactly what was said, without modifying or correcting the language (“I don’t know” will be transcribed exactly as is), as well as noting the non-verbal (e.g. laughter, pause, etc.).
- word-for-word transcription puts in writing everything that is said verbally. The final document will include even irrelevant content (“is this OK, can everybody hear me?”), repetitions (“I think we can get started. OK, let’s get started, then.”). Only hesitations (“um…”) will be omitted.
In addition, errors, spelling and syntax will be improved (“I dunno” will be turned into “I don’t know”).
- transcript writing, of the kind Ubiqus IO offers, includes every idea enunciated, but omits unrelated content, unnecessary repetitions and mistakes. The English (spelling and syntax) will also be improved should any errors have been made (see an example of our transcript-writing offers here).
Automatic transcription, myth or reality
The machine has yet to reach perfection
In an era where everyone has artificial intelligence on the tongue, our student and researcher clients are often surprised to read that, to date, there is no 100% reliable automatic transcription method.
Today, the best way to transcribe a recording is to do it “by hand and ear”. A writer listens to the audio and simultaneously types the transcription on the keyboard. This process, requiring precision and dexterity, will take between 4 and 6 hours per hour of audio.
However, this does not mean that there is no way of facilitating the transcription task: below, we list a few tools and ideas that can help.
Transcribing one hour of recorded interview will take between 4 and 6 hours*.
* for a relatively experienced person. It will require even more time of a beginner.
1. Automatic text transcription software
One relatively little-known aspect of YouTube is that it offers automatic subtitling for all videos.
Like its translation equivalent, GoogleTranslate, the automatic transcription tool produces decent output and goes a good way to fulfilling a need. However, it will only be of actual quality if the sound recording is very good or if only one person speaks at a time.
Even then, the transcription will be rough and require careful proofreading. For instance, it will contain no punctuation. All the grammar, spelling, proper nouns and acronyms will also need to be checked.
Use the automatic subtitling function
To get the transcript (via subtitles) of a video, you can create a “private” account on YouTube and enable automatic video transcription (in the subtitles tab). You can then amend YouTube’s “rough draft” and modify the transcript.
Be careful when retrieving the transcript, however. You will not be able to export it, except in sub-title formats (extensions .sbv; .srt; .vtt). Afterwards, it can be opened using an editor. Note that, in the resulting file, “time codes” will have been created, marking the time division necessary for sub-titling (but which you will have to remove to come out with a “clean” final transcript).
Sample .srt file with time codes:
Let’s get started. Well, hello there.
So, my name is Philip.
We are going to spend this next hour…
This is a commercial dictation product. The software can be trained to recognise one voice: yours — and is thus cannot be used to transcribe a conversation or meeting. The only way to deploy the software as a transcription tool would be to use the technique referred to as “parroting”.
Wearing a headset, you will need to repeat what you hear in your headset into the microphone for the software. Having tested it ourselves, we can safely say that it is not as easy as it sounds and won’t spare you the time required to proofread the software’s prose.
1.3. Automatic transcription software by Ubiqus IO
For high-quality interviews, Ubiqus IO offers its innovative Revised Automatic Transcription Solution.
Benefit : Revised Automatic Transcription is 30% less expensive than a standard transcript!
Thanks to the progress achieved with our artificial intelligence research, Ubiqus IO now offers an even more affordable transcription solution: an automatic transcript produced by our “in-house” software, and revised by one of our professional copy editors.
True to our promise to always deliver a ready-to-use product to our customers, we do not deliver raw transcription (as produced by our automatic transcription software). This first draft is always re-processed by a professional copy editor; however, using the machine for part of the task makes it possible to lower costs… by around 30%.
Send it to us for analysis via this form.
2. Software that plays audio, later transcribed “by hand”
Most free transcription software does not provide automatic audio to text transcription. In general, this term refers to software that allows audio playback and simultaneous capture, and thus offers no other assistance than having the two tools combined in one.
- For example, our professional copy-editing colleagues use Express Scribe – however this is pay software.
- A small, smart and free online solution is also available, by the name of otranscribe. This site allows you to play audio easily (by modulating the listening speed) all the while typing what you hear into a small text editor. Simple and easy to use.
How long does it take to transcribe an interview?
Using only audio playback software paired with a publisher such as otranscribe, transcribing one hour of interview content will take a minimum of 4 hours for a professional to six hours for a novice, or sometimes even twice that much if the subject is technical and the recording of poor quality.
The use of automatic transcription software (such as that which Ubiqus IO uses) reduces this time by a quarter or at best by a third, if and only if the audio is of excellent quality.
An amateur will dedicate between half a day and a full day to transcribing a single hour of audio!
As you will have understood, if you are not a transcription professional, it will take you anywhere from a half to a whole day to transcribe a simple hour of audio, subject to:
- audio quality;
- your typing speed;
- your spelling, syntax and grammar;
- your knowledge of the subject matter (checking proper nouns, acronyms, technical jargon).
Just imagine if you have several hours of interview to transcribe!
Incorporating transcribed interviews into your thesis or dissertation
Once your interviews have been transcribed by you or by a service provider, you will need to incorporate them into your final dissertation. I recommend that you read the excellent blog post written by our friends at Scribbr which will enlighten you on this point, as well as on their thesis review and correction services.
To sum up:
6 : the minimum number of hours required by a non-professional to transcribe one hour of audio provided good quality;
20 : the number of pages in a transcript of one hour of audio;
110 : the price (in € excluding VAT) of a transcript produced by Ubiqus IO using Revised Automatic Transcription, for one hour of interview audio..
So, have you decided to save time, and a receive an end-product that can be used and integrated directly into your dissertation? Click here for your instant, personalised quote.