A few issues raised by Anton

[danyaljj]

I though the following information might be helpful to you as you make final changes to the dataset. It seems that a few claims have a subset of equivalent perspectives with different ids but the text is the same. The claim ids are:
- dev -- 542, 469
- train -- 547, 795, 247, 241 
- test -- none

For example, for claim 542, one set of equivalent perspectives is:
pId 1776: 'The state benefits from the skills of a university educated populace '
pId 3986: 'The state benefits from the skills of a university educated populace '
pId 24372: 'An educated population is for the benefit of the state'
pId 22509: 'The state is benefited by those with university educated skills.'

Clarify the notation:

what is the difference between stance_label_3 and stance_label_5?

which means:

We following two labeling conventions: 
 - 3 labels ("stance_label_3"): Support, Undermine, Not-a-valid-perspective; 
 - 5 labels ("stance_label_5"): Support, Mildly-support, Undermine, Mildly-undermine, Not-a-valid-perspective;