Generally, a document in the real world (especially in the news domain) contains multiple concepts, so multi-label text classification is a very important technology for automatic metadata ...