Discipline-Independent Text Information Extraction From Heterogeneous Styled References Using Knowledge From The Web