Automatic extraction of illocutionary force types in a dialogue, based on context and modal.