LateInteractionEmbedder: Late Interaction Embedder (Simplified ColBERT-style)

In VectrixDB: Lightweight Vector Database with Embedded Machine Learning Models

Late Interaction Embedder (Simplified ColBERT-style)

Description

Token-level embeddings for late interaction scoring

Public fields

dimension

Token embedding dimension

language

Language setting ("en" or "ml")

Methods

Public methods

• LateInteractionEmbedder$new()

• LateInteractionEmbedder$embed()

• LateInteractionEmbedder$score()

• LateInteractionEmbedder$clone()

---

Method new()

Create a new LateInteractionEmbedder

Usage

LateInteractionEmbedder$new(dimension = 64, language = "en")

Arguments

dimension

Embedding dimension per token

language

Language behavior ("en" = ASCII-focused, "ml" = Unicode-aware)

---

Method embed()

Embed texts to token-level embeddings

Usage

LateInteractionEmbedder$embed(texts)

Arguments

texts

Character vector of texts

Returns

List of matrices (each matrix is token embeddings for a document)

---

Method score()

Compute late interaction (MaxSim) score

Usage

LateInteractionEmbedder$score(query_embeddings, doc_embeddings)

Arguments

query_embeddings

Query token embeddings matrix

doc_embeddings

Document token embeddings matrix

Returns

Numeric score

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LateInteractionEmbedder$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

VectrixDB documentation built on Feb. 20, 2026, 5:09 p.m.