Flair for NLP: why MIT movie corpus has extra labels than original dataset?

Question

I have a question related to flair framework for natural language processing: I am trying to load simple MIT movie corpus for NER task. However, I the output co