Text postprocessing

Then someone must correct the text and make sure that what is said is what is written (and eventually adjust the written text to follow the set) (speaker) (xx months)

• The read texts must be edited to match the audio as accurately as possible. Also light mistakes, corrections, repetitions are kept and written out in the text transcript.
• It is very important to write out the numbers, abbreviations, acronyms and any pronunciation of special characters (like @, /, - etc.). People tend to read these unsystematically and the actual pronunciation of these may vary.
• Often, another round of audio cleaning will be also done at this point to remove all remaining coughs, mistakes that weren’t noticed before.

Text processing / normalisation

Using rule-based technologies (Sjur writes this)

• normalising acronyms and abbreviations
• normalising digits of various kinds
• exceptional pronunciation
• do you need fonemic / non-orthographic text?