Enum ExtractionMethod 

#[non_exhaustive]
pub enum ExtractionMethod {
    Pattern,
    Neural,
    Lexicon,
    SoftLexicon,
    GatedEnsemble,
    Consensus,
    Heuristic,
    Unknown,
    Rule,
    ML,
    Ensemble,
}

Extraction method used to identify an entity.

Research Context

Different extraction methods have different strengths:

Method	Precision	Recall	Generalization	Use Case
Pattern	Very High	Low	N/A (format-based)	Dates, emails, money
Neural	High	High	Good	General NER
Lexicon	Very High	Low	None	Closed-domain entities
SoftLexicon	Medium	High	Good for rare types	Low-resource NER
GatedEnsemble	Highest	Highest	Contextual	Short texts, domain shift

See docs/ for repo-local notes and entry points.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Pattern

Regex pattern matching (high precision for structured data like dates, money). Does not generalize - only detects format-based entities.

Neural

Neural model inference (BERT, GLiNER, etc.). The recommended default for general NER. Generalizes to unseen entities.

Lexicon

👎Deprecated since 0.2.0: Use Neural or GatedEnsemble instead

Exact lexicon/gazetteer lookup (deprecated approach). High precision on known entities, zero recall on novel entities. Only use for closed domains (stock tickers, medical codes).

SoftLexicon

Embedding-based soft lexicon matching. Useful for low-resource languages and rare entity types. See: Rijhwani et al. (2020) “Soft Gazetteers for Low-Resource NER”

GatedEnsemble

Gated ensemble: neural + lexicon with learned weighting. Model learns when to trust lexicon vs. context. See: Nie et al. (2021) “GEMNET: Effective Gated Gazetteer Representations”

Consensus

Multiple methods agreed on this entity (high confidence).

Heuristic

Heuristic-based extraction (capitalization, word shape, context). Used by heuristic backends that don’t use neural models.

Unknown

Unknown or unspecified extraction method.

Rule

👎Deprecated since 0.2.0: Use Heuristic or Pattern instead

Legacy rule-based extraction (for backward compatibility).

ML

👎Deprecated since 0.2.0: Use Neural instead

Legacy alias for Neural (for backward compatibility).

Ensemble

👎Deprecated since 0.2.0: Use Consensus instead

Legacy alias for Consensus (for backward compatibility).

Implementations

impl ExtractionMethod

pub const fn is_calibrated(&self) -> bool

Returns true if this extraction method produces probabilistically calibrated confidence scores suitable for calibration analysis (ECE, Brier score, etc.).

Calibrated Methods

• Neural: Softmax outputs are intended to be probabilistic (though may need temperature scaling for true calibration)
• GatedEnsemble: Produces learned probability estimates
• SoftLexicon: Embedding similarity is pseudo-probabilistic

Uncalibrated Methods

• Pattern: Binary (match/no-match); confidence is typically hardcoded
• Heuristic: Arbitrary scores from hand-crafted rules
• Lexicon: Binary exact match
• Consensus: Agreement count, not a probability

Example

use anno_core::ExtractionMethod;

assert!(ExtractionMethod::Neural.is_calibrated());
assert!(!ExtractionMethod::Pattern.is_calibrated());
assert!(!ExtractionMethod::Heuristic.is_calibrated());

pub const fn confidence_interpretation(&self) -> &'static str

Returns the confidence interpretation for this extraction method.

This helps users understand what the confidence score means:

• "probability": Score approximates P(correct)
• "heuristic_score": Score is a non-probabilistic quality measure
• "binary": Score is 0 or 1 (or a fixed value for matches)

Trait Implementations

impl Clone for ExtractionMethod

fn clone(&self) -> ExtractionMethod

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ExtractionMethod

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for ExtractionMethod

fn default() -> ExtractionMethod

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ExtractionMethod

fn deserialize<__D>( __deserializer: __D, ) -> Result<ExtractionMethod, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for ExtractionMethod

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Hash for ExtractionMethod

fn hash<__H>(&self, state: &mut __H)

where __H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for ExtractionMethod

fn eq(&self, other: &ExtractionMethod) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for ExtractionMethod

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for ExtractionMethod

impl Eq for ExtractionMethod

impl StructuralPartialEq for ExtractionMethod