Module types 

Evaluation types: MetricValue, GoalCheckResult, etc.

These are shared primitives for evaluation that can be reused across NER evaluation and other evaluation tasks.

Structs

CorefChainStats

Statistics for stratified coreference evaluation.

CorefDocStats

Document-level statistics for coreference evaluation.

GoalCheck

Individual goal check.

GoalCheckResult

Result of checking evaluation goals.

LabelShift

Label shift between training and evaluation entity types.

MetricDivergence

Divergence between coreference metrics.

MetricValue

A type-safe metric value bounded to [0.0, 1.0].

MetricWithVariance

A metric with variance and confidence interval.

Enums

DocumentScale

Document scale classification based on token count.