Develope a module to use token from my own dictionary!

[du...@gmail.com]

Hi all!
I'm going to develope a system to filter spam in Vietnamese language base on the semantic of token(or term). This term is retrieved from my own dictionary. I tend to use some modules of Spamassassin such as sa-learn but my problem is :
 Don't know exactly where  does code to retrieve token from mail( in training corpus) locate? 
 So is there any one know or have done something similar , please help me!
Thanks in advance!
Your sincerely!
Le Huy Duc