Struct DocumentRanker 

pub struct DocumentRanker { /* private fields */ }

Multi-factor document ranker combining BM25 lexical scoring with semantic similarity.

Usage

use ipfrs_semantic::document_ranker::{DocumentRanker, RankingConfig, DocumentIndex};

let config = RankingConfig::default();
let mut ranker = DocumentRanker::new(config);

let doc = DocumentIndex::from_tokens("doc1", &["hello", "world"], None);
ranker.index_document(doc);

let results = ranker.rank(&["hello".to_string()], None);
assert!(!results.is_empty());

Implementations

impl DocumentRanker

pub fn new(config: RankingConfig) -> Self

Creates a new DocumentRanker with the given configuration.

pub fn index_document(&mut self, doc: DocumentIndex)

Indexes (or re-indexes) a document.

Inserting a document with the same doc_id as an existing one will overwrite the previous entry. After insertion the average document length and IDF cache are refreshed for all terms present in the new document.

pub fn rank( &mut self, query_terms: &[String], query_embedding: Option<&[f64]>, ) -> Vec<RankedDocument>

Ranks all indexed documents against the given query terms and optional query embedding.

Results are filtered by RankingConfig::min_score, sorted by descending combined score, and truncated to at most RankingConfig::max_results entries. Each returned RankedDocument carries a 1-based rank field.

pub fn bm25_score(&self, doc: &DocumentIndex, query_terms: &[String]) -> f64

Computes the BM25 score for a single document given the query terms.

Uses the Robertson–Sparck Jones IDF with BM25+ numerator adjustment.

pub fn compute_idf(&self, term: &str) -> f64

Computes the IDF of a term using Robertson–Sparck Jones smoothed formula:

ln((N - df + 0.5) / (df + 0.5) + 1)

where N is the total number of indexed documents and df is the document frequency of term.

pub fn cosine_similarity(a: &[f64], b: &[f64]) -> f64

Computes the cosine similarity between two embedding vectors.

Returns 0.0 when either vector is zero-length or the lengths differ.

pub fn update_avg_length(&mut self)

Recomputes the average document length across all indexed documents.

Called automatically after each index_document.

pub fn update_idf_cache(&mut self, terms: &[String])

Refreshes the IDF cache for the given term list.

Existing cache entries for terms not in terms are preserved.

pub fn document_count(&self) -> usize

Returns the number of documents currently in the index.

pub fn stats(&self) -> &RankerStats

Returns a reference to the accumulated query statistics.

pub fn avg_doc_length(&self) -> f64

Returns the current average document length used by BM25.

pub fn get_document(&self, doc_id: &str) -> Option<&DocumentIndex>

Returns a reference to a specific indexed document, if present.