Speech

Data File: Speech.mul

This file has no indexer. Directly after one chunk ends, another begins.

Chunks:

short keywordID; // Note: This is big-endian
short keywordLength; // Note: This is also big-endian
char[keywordLength] keyword;

When the client says a certain keyword, it only has to send the keyword ID to the server.
As keywords can be in different languages, the server no longer needs to translate speech.