Contemporary history of Luxembourg

Archiving and Oral History. Hands-On history: automated transcription.

In the context of the project “24.02.22, 5 am: Testimonies from the War”, one of the tasks in the process of creating a digital archive of testimonies is to transform the audio testimony into a transcript. We have important questions to consider when transcribing :Which way and type of tranсribing corresponds to our purpose? Which type of transcript will be most useful for your particular context?
Ways to Transcribe Interviews
Transcription can be done in several different ways: In the scientific literature, four main ways of transcription are distinguished.
Using:Audio-to -Text Conversation; Transcribing manually; Outsourcing to a Transcription Agency; Hiring a Freelance Transcriptionist.
Іn our project “24.02.22, 5 am: Testimonies from the War”, after comparing manual transcription with automatic, we decided to use automatic transcription.
In recent years, enormous technological advances have been made in the field of speech recognition. In the scientific literature, we can find many publications summarizing the experience of projects that solve the problem development of speech recognition and speech analysis technologies in the context of oral history. Despite the improvements in speech recognition, the transcripts contain recognition errors.
At the initial stage, Audio-to-Text Conversation did not recognize the Ukrainian language well. Especially when a speaker names a person, an organization, an acronym, name of the area, a word that was not included in the program, ect.
How we can contribute to the improvement of the program?
Just actively working with additional features of the program. At first glance, it requires more time. But this investment pays off. The artificial intelligence of the program will take into account its mistake next time and improve the quality itself.
The next time you upload a file, simply select My Vocabulary to ensure that the words you have added to your vocabulary are identified and transcribed with even greater precision.
Choosing Audio-to-Text Conversation we contribute to the modernization of the transcription process.

Show this publication on our institutional repository (orbi.lu).