transkribering

Transkribering is the process of converting spoken language into written text. It applies to audio or video material and serves purposes in research, journalism, media production, education and legal contexts. Transcriptions can vary in level of detail, depending on the goals of the project.

Two common approaches are verbatim and clean read. Verbatim transcription aims to capture everything spoken, including hesitations, repetitions, false starts, pauses, fillers, and overlapping speech. Clean or intelligent verbatim suppresses or simplifies some of these features to produce a more readable text. Transcripts may also include time stamps and speaker labels to indicate when each person spoke or when a particular event occurred.

Notational conventions often cover time codes, speaker identification, and notation of non-verbal events (laughter, coughing, ambient noise). Many workflows also include audio refinement, segmentation into utterances, and quality control through proofreading or double transcription. Automated speech recognition (ASR) tools can speed up the process, but human review is typically necessary to reach high accuracy, especially for specialized terminology or noisy recordings.

Common applications include linguistic and social research, documentary or media subtitling and captioning, interviews and oral histories, and, in some jurisdictions, legal transcription. The quality of a transcript is usually measured by accuracy against the original audio and by its usefulness for the intended task, including proper time alignment and clear speaker attribution. Ethical considerations, such as consent and data anonymization, are often important in transkribering projects.