I need to transcribe my interview recordings. Does SLU offer or recommend any particular software for this?

In short

• SLU does not have a university-wide license for automatic transcription, but you can check whether your department has purchased any software for this purpose.
• Before using or purchasing a transcription service, see information below about what needs to be considered when choosing a software.

Transcription options

There are different kinds of tools and services that you can use to transcribe audio files. Some software allows you to play the sound file while you type the transcription manually (these tools have keyboard shortcuts for repeating, playing slowly, etc), and more advanced ones that will automatically create a transcription based on your audio file. These usually come with a subscription fee, or charge you based on how much you transcribe.

Automatic transcription is not always the most convenient or fastest option. How well the tool performs depends on a number of factors, not only sound quality but also language, accent, and subject area. You may find that you need to spend a lot of time editing and correcting the transcriptions to ensure they are accurate, and that doing the transcription manually is a viable option.

If you prefer manual transcription but do not want to do it yourself, it is possible to use the framework agreement for language services. Manual transcription is available as an add-on service under this agreement. You can read about it here:

• Ramavtalet för språktjänster (Swedish)

You can find more information about the agreement here: 

• Alla ramavtal (Swedish)
   

What do I need to consider regarding personal and sensitive data?

You need to consider what type of data are in the interviews that will be transcribed. Your interview material most likely constitutes personal data, even though you may have avoided recording any names. A voice can also be personal data. When personal data are processed the GDPR (General Data Protection Regulation) needs to be considered.  

• Sensitive personal data | IMY

Many transcription tools are cloud based and owned by American companies. If such software is used, it means that personal data will be transferred to a third country, since the US is outside the EU/EES. If you are considering using transcription software that transfers the data to a third country, for example, the US, you need to make sure that the transfer is legal. You can read more about how to do that here:

• Transferring personal data to non-EU countries (in particular the US) | Medarbetarwebben.

Keep in mind that SLU needs a data processing agreement with the provider of the software if the interview includes personal data. This is because the provider will the process personal data for which SLU is responsible. A data processing agreement is not needed if the personal data are not transferred outside the organization and is handled only locally on your computer by you.

Sometimes the interview may also contain sensitive personal data and/or classified information. SLU has a policy stating that sensitive personal data and classified information must not be stored in a cloud. This means that you should not use transcription services with cloud storage if your interview recordings contain sensitive personal data or classified information.
 

Can I use an AI service to transcribe my interview?

Many transcriptions software today have AI components or are entirely AI services. Many of these AI systems use the data they receive to learn, in other words, the company uses the data to train its AI. This most likely means that the data is spread or leaked to others. If the interview contains personal data or classified information you should not use transcription software with AI that learns from the data. The reason for this is that the processing would not comply with the law.  Therefore, you should always investigate whether the transcription software uses AI before transferring any data. If it does, you may only use the service if you can ensure that the data is stored locally on your computer and not uploaded online when using the service or that the data isn´t used for training the AI.

You can find more information and advice on using AI services here:

• Advice on using AI services at work | Medarbetarwebben
   

Can SLU give any examples of specific transcription software?

We can give you some examples of tools but please note that we do not provide technical support. 

An example of an AI service that can be downloaded and installed locally is

• Whisper.cpp

The whisper.cpp program is designed to run locally without uploading anything to the cloud. Although OpenAI originally developed Whisper, this version does not run in their (Microsoft's) cloud, and the whisper.cpp implementation was developed independently. Whisper.cpp requires you to build the transcription software yourself. This version is easier to run locally:

• Local Transcribe with Whisper

If you’re using a Mac, MacWhisper may be an easier alternative.

• MacWhisper

There is also a system called Ragnar. In order for Ragnar to be used, you need to make sure it is only accessible from your local machine. Ragnar also requires some command line skills to get started.

• Ragnar

Another service called aTrain should work on Windows, can be installed via Microsoft Store, and also fully locally:

• aTrain

Microsoft Word and Teams include transcription functions, but sensitive personal data and classified information should not be handled in Office 365.

• Office 365

Whichever one you choose, you must still follow SLU advice for using AI services and keep in mind what type of data you have in the interview. To be absolutely certain that nothing leaks, you should restrict the software’s network access, for example, via a firewall. This applies to any software that can access your data. Contact IT support for help with firewall settings.