I made a thread a few days back where I tried to the same myself.
The answer is simple: it's hardcoded.
The longer answer is: leader music is coded into the leader, and leader speech is coded into the scene. Therefore, Tokugawa who uses the default oda scene speaks correctly, while Oda Nobunaga who uses a different oda scene has the japanese leader music playing. But I can't define who gets what (otherwise I'd have custom music already).