These are pronunciation rules and exceptions lists for each language.

use:
  speak --compile=<voicename>

to compile the data files for each language in espeak-data (eg. en_dict).

