It is important to distinguish between spam policies and content filtering policies. Brightmail Gateway will NOT identify why a given message was flagged as spam. This is by design, in order to avoid abuse by spammers looking for ways to work around Symantec's spam filters.
The only policies that use the dictionaries are the content filtering policies that are deployed locally by an email administrator. These will be tracked separately from the spam verdict. Today, the Message Audit Log will identify which policy fired, but not which word within a dictionary triggered the verdict. As Cricket17 indicates, there is a proposed enhancement in the Ideas section of the forums, and we are actively investigating this as an enhancement for a future release.