Orthographic Transcription

The most basic type of annotation that makes a collection of speech recordings into a speech corpus is some kind of orthographic transcription. This can range from a simple chain of words per recording item (based for instance on the script that was used during the recording) to an extensive labeling of several different semantic layers^8.5. The choice about what is to be included in the transcript is dependent on the type of speech corpus and the intended usage. For example, a corpus of read speech items over the telephone network with the aim to train automatic speech recognition algorithms does not need any elaborated labeling of discourse events. A corpus containing dialogue speech between two or more persons that is subject to scientific investigations will require much more effort.