Struct EnsembleNER 

pub struct EnsembleNER { /* private fields */ }

Ensemble NER that runs ALL backends and resolves conflicts via weighted voting.

Unlike StackedNER (priority-based cascade), EnsembleNER:

1. Runs ALL backends in parallel (conceptually)
2. Groups overlapping spans into conflict clusters
3. Resolves via weighted voting with type-conditioned weights
4. Applies agreement bonus when multiple backends agree

When to Use

• EnsembleNER: Maximum accuracy, latency not critical
• StackedNER: Production, predictable latency, early exit

Example

use anno::{EnsembleNER, Model, RegexNER, HeuristicNER};

// Default: uses all available backends
let ensemble = EnsembleNER::new();

// Custom: specific backends
let custom = EnsembleNER::with_backends(vec![
    Box::new(RegexNER::new()),
    Box::new(HeuristicNER::new()),
]);

let entities = custom.extract_entities("Contact us at test@example.com", None)?;

Algorithm

1. Collect candidates: Run each backend, tag results with provenance
2. Cluster overlaps: Group candidates with >50% span overlap
3. Weighted vote: Score each candidate by backend_weight * confidence
4. Agreement bonus: Add +0.10 when 2+ backends agree on type
5. Select winner: Highest weighted score wins the cluster

Implementations

impl EnsembleNER

pub fn new() -> Self

Create ensemble with all available backends.

pub fn with_backends(backends: Vec<Box<dyn Model + Send + Sync>>) -> Self

Create with custom backends.

pub fn with_weights(self, weights: HashMap<String, BackendWeight>) -> Self

Set custom backend weights.

pub fn with_agreement_bonus(self, bonus: f64) -> Self

Set the agreement bonus (added when multiple backends agree).

pub fn with_min_confidence(self, min: f64) -> Self

Set minimum confidence threshold.

Trait Implementations

impl BatchCapable for EnsembleNER

fn optimal_batch_size(&self) -> Option<usize>

Get the optimal batch size for this model, if applicable.

fn extract_entities_batch( &self, texts: &[&str], language: Option<&str>, ) -> Result<Vec<Vec<Entity>>>

Extract entities from multiple texts in a batch.

impl Default for EnsembleNER

fn default() -> Self

Returns the “default value” for a type.

impl Model for EnsembleNER

fn extract_entities( &self, text: &str, language: Option<&str>, ) -> Result<Vec<Entity>>

Extract entities from text.

fn supported_types(&self) -> Vec<EntityType>

Get supported entity types.

fn is_available(&self) -> bool

Check if model is available and ready.

fn name(&self) -> &'static str

Get the model name/identifier.

fn description(&self) -> &'static str

Get a description of the model.

fn capabilities(&self) -> ModelCapabilities

Get capability summary for this model.

fn version(&self) -> String

Get a version identifier for the model configuration/weights.

impl StreamingCapable for EnsembleNER

fn recommended_chunk_size(&self) -> usize

Get the recommended chunk size for streaming extraction.

fn extract_entities_streaming( &self, chunk: &str, offset: usize, ) -> Result<Vec<Entity>>

Extract entities from a chunk of text, adjusting offsets by the chunk’s position.

impl NamedEntityCapable for EnsembleNER