agrag.eval.extraction.RelationBreakdown
Bases: TypedDict
The score_breakdown of a relation quality metric for one case.
The label lists are 0/1 values over the union of gold and predicted items:
y_true marks gold items and y_pred marks predicted ones. The
relaxed_ lists use an overlap of at least 0.5 to align entities.
Attributes:
- kind (
Literal['entity', 'relation']) – - relaxed_y_pred (
list[int]) – - relaxed_y_true (
list[int]) – - y_pred (
list[int]) – - y_true (
list[int]) –
kind
kind: Literal['entity', 'relation']
relaxed_y_pred
relaxed_y_pred: list[int]
relaxed_y_true
relaxed_y_true: list[int]
y_pred
y_pred: list[int]
y_true
y_true: list[int]