NLP Subdocument classification

Can anyone make any recommendations on how to implement the following classification for subdocuments with inconsistent line spacing?