$\bigcirc$ Annotation and lexicon contents

If not already done in the previous steps^5.4, write a simple script to extract labels from the annotation files and check them for inconsistencies.

• Cross-check the found labels with the documentation of the labeling. Are all found labels documented? Are there any documented labels not found in the annotations?
• Report any digits or numerals that are not written in their full orthographic form.
• Report any punctuation used in the annotations. There shouldn't be any except in cases where they are separated from other items by white space and have a special meaning (for instance prosodic).
• Report any words that are written with an initial capital because they are at the beginning of a sentence.
• Cross-check all words extracted from the transcripts with the spelling in the orthographic part of the lexicon.

Also you might check the timing information in label files for overlapping segments or gaps between segments, if this should not happen according to your reference.