Although previous methods have advanced text structure parsing effectively, they face challenges such as not leveraging the abundance of unlabelled data and focusing mainly on content-inferred ...