Transcription

In January (2021), Microsoft launched its own automatic transcription function through MS Word. The feature is available through their MS365 solution for users with a valid Microsoft license when logging in to the web version. 

Traditionally, the IT department has not been in the habit of writing or marketing anything about specific possibilities in either the Office programs or other application solutions, but since transcription is so relevant for many users at OsloMet and that this is a function with a special privacy focus, we think it is both important and correct that this functionality is communicated in the right way. 

The automatic transcription feature in Word is only available in the web version which is accessed by logging in to OsloMet’s MS365 (formerly O365). 

1. There are basically two ways in which the automatic transcription feature can be used. You can submit a finished audio recording (audio file with recording of interview / conversation). 
   
   or
    
2. Start transcribing using a live recording. The latter is not recommended because you do not have control over what is said in the audio recording. Alternatively, the live recording feature can be used when playing a finished audio recording (see separate routine for students below). 

As long as the transcription feature in MS Word is part of MS Office, Microsoft’s cloud solution applies the same principles for data classification as for Microsoft otherwise, which means that you can only use the solution for yellow data. That is, general personal information, provided that you have the prior consent of the participants. 

What about red (sensitive/confidential) personal data? 

Automatic transcription of red data is not approved. This type of processing must take place on a system that is approved for processing and storing red data.

To prevent sensitive/confidential material from being used in a service, you must therefore be familiar with the contents of the audio file before sending it for transcription. 

Information for participants 

It will be a good practise that the transcription method has been informed to the participants in the study if you are to use this in research

Uploading the audio file 

The audio file to be retrieved for transcription must be available from your PC, preferably in a storage location in OneDrive. 

When the program reads the audio file, the audio data is sent to a program at Microsoft’s (“speech-to-text") which, with the help of artificial intelligence, has learned to understand speech and translate this into text. 

The transcribed material is not stored at Microsoft, but is sent directly to your PC when it has been transcribed, where you have the opportunity to download the text file. 

Keep track of your files 

We recommend that you be careful about file placement and that such data files are not accidentally downloaded or copied to different locations, but that this is included in your workbook (in Teams or OneDrive) which is guaranteed deletion when the project is completed or when material is not needed anymore. The same goes for the audio file. 

How to bypass storage on private equipment?

It is not allowed to download an audio file with personal data on a private machine, so special routines must be followed. 

If you have a finished audio recording in Web form, read through the Dictaphone app, the solution is that you can play it in a browser at the same time as you use the recording function in Word (online). As if you were simulating a live recording. 

Where can I find the transcribed file? 

When the interview is finished playing, an audio file is created which is placed on your OneDrive folder under the \Transcribed Files folder. 

Remember to make sure that there is no sensitive/confidential data in the recording before you start transcribing and that you do not sync down to a private machine but leave it in O365-online. 

What about the Norwegian language? 

The program handles several languages. Although it is best “trained” in English, the tests we have done in Norwegian show it works quite well. 

Word versus NVivo and others 

Tests performed at OsloMet show that the quality of transcription in Word is superior the quality that came out of Google and NVivo’s transcription results on the same text base, respectively. 

What if more people talk? 

The program also handles that more people speak in groups, but the more people in dialogue, the more demanding it is of course for the program. And the more dialects there are, the harder it is to understand. We recommend testing the feature somewhat before it is decided to be used. If the quality is just that right, the time you need to rewrite material may take as long as if you had transcribed everything manually from the start. 

Timestamp 

The text file is indicated with a time stamp to be able to follow the time through the conversation, who said what when. 

Microsoft has so far only provided 300 minutes (5 hours) per month per user and this restriction cannot be extended as of today.

Providers of other transcription services 

There are several other transcription services ala Microsoft Word on the market, but so far OsloMet has not entered into an agreement with any other suppliers. 

Some companies will sell their transcription service under their own name; however, the provider of the actual transcription may be provided by either Google or Microsoft. For example, Benevis, which uses Google’s transcription feature. 

A third-party supplier, Speechmatics, sells its product through what many know as NVivo Transcript. 

How much does it cost? 

Both Benevis and NVivo Transcript take a cost in that you must buy a certain number of transcription minutes. MS Word, on the other hand, is free until further notice.