Objective: Identify the human values expressed by each actor in a factual news article, aiming to cover all actors appearing in its subevents. For Level-1 values, assign a direction (aligned or contradictory); for Level-2 values, assign labels without direction.
Provided data:
guid: article id
content: article content
List of main actors
Level-1/2 human value categories
Output:
Task-1a: <guid, actor, level-2 value>
Task-1b: <guid, actor, level-1 value, Aligned |Contradictory>
Error definition:
Omission error: missed gold labels
Reversal error : predicted opposite direction (aligned ↔ contradictory)
Objective: Identify, for each subevent, the human values expressed by each actor: (1) assign Level-1 value labels with a direction (aligned or contradictory), and (2) assign Level-2 value labels without direction, aiming to cover all actors present in the subevent.
Provided data:
guid: article id
content: article content
subevent id
subevent content
List of main actors
Level-1/2 human value categories
Output:
Task-2a: <guid, subevent_id, actor, level-2 value label>
Task-2b: <guid, subevent_id, actor, level-1 value label, Aligned |Contradictory>
Error definition:
Omission error: missed gold labels
Reversal error : predicted opposite direction (aligned ↔ contradictory)