Struct Scorer

pub struct Scorer<K, D = u32> { /* private fields */ }

Efficiently scores the relevance of a query embedding to document embeddings using BM25. K is the type of the document id and D is the type of the token embedder.

Implementations

impl<K, D> Scorer<K, D>

where D: Eq + Hash + Clone, K: Eq + Hash + Clone,

pub fn new() -> Scorer<K, D>

Creates a new Scorer.

pub fn upsert(&mut self, document_id: &K, embedding: Embedding<D>)

Upserts a document embedding into the scorer. If an embedding with the same id already exists, it will be replaced. Note that upserting a document will change the true value of avgdl. The more avgdl drifts from its true value, the less accurate the BM25 scores will be.

pub fn remove(&mut self, document_id: &K)

Removes a document embedding from the scorer if it exists.

pub fn score( &self, document_id: &K, query_embedding: &Embedding<D>, ) -> Option<f32>

Scores the embedding for the given document against a given query embedding. Returns None if the document does not exist in the scorer.

pub fn matches(&self, query_embedding: &Embedding<D>) -> Vec<ScoredDocument<K>>

Returns all documents relevant (i.e., score > 0) to the given query embedding, sorted by relevance.

Trait Implementations

impl<K: Default, D: Default> Default for Scorer<K, D>

fn default() -> Scorer<K, D>

Returns the “default value” for a type.