Module error_analysis 

Error analysis for NER systems.

Categorizes and analyzes prediction errors to guide improvement efforts.

Error Categories

• Boundary Errors: Correct type but wrong span
• Type Errors: Correct span but wrong type
• False Positives: Predicted entity that doesn’t exist
• False Negatives: Missed a real entity

Example

use anno_eval::eval::error_analysis::{ErrorAnalyzer, PredictedEntity};
use anno_eval::eval::datasets::GoldEntity;
use anno::EntityType;

let analyzer = ErrorAnalyzer::default();

let predictions = vec![
    PredictedEntity::new("John", "PER", 0, 4),
    PredictedEntity::new("Google", "LOC", 14, 20),  // Wrong type!
];

let gold = vec![
    GoldEntity::with_span("John Smith", EntityType::Person, 0, 10),    // Boundary error
    GoldEntity::with_span("Google", EntityType::Organization, 14, 20), // Type error
];

let report = analyzer.analyze(&predictions, &gold);
println!("Boundary errors: {}", report.boundary_errors.len());
println!("Type errors: {}", report.type_errors.len());

Structs

EntityInfo

Entity information for error reports.

ErrorAnalyzer

Analyzer for NER prediction errors.

ErrorInstance

A specific error instance.

ErrorPattern

Common error pattern.

ErrorReport

Comprehensive error analysis report.

PredictedEntity

A predicted entity for error analysis.

TypeErrorStats

Error statistics for a specific entity type.

Enums

ErrorCategory

Error category for NER analysis.