The rules

The rules listed below were taken from Kirchhoff's ``Two-level Modelling of Speech Variant Rules'' ([Kir95]). The numbering of the rules was maintained in order to keep the reference to the definitions of the rules simple. In the cited report we find detailed information about the individual rules as well as some words about the motivation for them. Any change in the present implementaion of the rules will be explicitly stated. The rules were implemented using the XFST-command read regex. This way, one regular expression had to be defined for each individual rule and subrule. The regular expressions defined are explicitly specified below.

Following the explanation of the implementation of each rule, information about its size (as a network) and storage is given. This information can be obtained using the XFSTcommands print size and print file-length.

The rules can be found in the directory: XFST/SOFTWARE/Rules/ in the Bielefeld VM archive.

 

• Naming Conventions for the Rules
• Definition of the phoneme classes
• The Implementation of the Rules

Thorsten Trippel Mon Mar 20 14:35:50 MET 2000