The default character encoding for XML documents is utf8, which has 7-bit ASCII as a proper subset. Character codes in VoiceXML documents that are greater than 127 will therefore be interpreted as the first byte of a multi-byte utf8 sequence. You can force the XML parser to use another codepage by specifying the desired encoding as the very first thing in the file:
<?xml version="1.0" encoding="GBK"?>