Struct LSHBlocker 

pub struct LSHBlocker {
    pub num_bands: usize,
    pub rows_per_band: usize,
    pub ngram_size: usize,
}

Locality-Sensitive Hashing (LSH) for blocking.

Blocking reduces the O(n²) pairwise comparison problem by grouping likely-coreferent mentions into “blocks” using hash signatures.

Algorithm

1. Compute a character n-gram signature for each mention
2. Hash the signature using multiple hash functions
3. Mentions with identical hash in any band are candidates

This achieves sub-quadratic scaling while maintaining high recall for truly coreferent pairs.

Fields

num_bands: usize

Number of hash bands

rows_per_band: usize

Rows per band

ngram_size: usize

N-gram size for signatures

Implementations

impl LSHBlocker

pub fn new(num_bands: usize, rows_per_band: usize) -> Self

Create a new LSH blocker.

pub fn candidate_pairs(&self, mentions: &[MentionRef]) -> Vec<(usize, usize)>

Compute candidate pairs for a set of mentions.

Returns pairs of indices (i, j) where i < j that are candidates for cross-document coreference.

pub fn signature_similarity(sig1: &[u64], sig2: &[u64]) -> f64

Estimate Jaccard similarity from minhash signatures.

Trait Implementations

impl Clone for LSHBlocker

fn clone(&self) -> LSHBlocker

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LSHBlocker

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

Formats the value using the given formatter.

impl Default for LSHBlocker

fn default() -> Self

Returns the “default value” for a type.