pub fn stepwise_error_analysis<E: ClusterEncoder + Clone, S: MergeScorer + Clone>(
benchmark: CrossContextBenchmark,
topics: &[Topic],
documents: &[LongDocument],
encoder: E,
scorer: S,
) -> Result<HashMap<StepwiseAnalysis, CrossContextEvalResults>>Expand description
Run stepwise error analysis as in xCoRe Table 5.
This helps identify which pipeline stage is the bottleneck:
- Mention extraction
- Within-context clustering
- Cross-context cluster merging