mmr_rerank

Maximal Marginal Relevance (MMR) reranking

internal

A lightweight vector database for text retrieval in R with embedded
machine learning models and no external API (Application Programming
Interface) keys. Supports dense and hybrid search, optional HNSW
(Hierarchical Navigable Small World) approximate nearest-neighbor indexing,
faceted filters with ACL (Access Control List) metadata, command-line
tools, and a local dashboard built with 'shiny'. The HNSW method is
described by Malkov and Yashunin (2018) <doi:10.1109/TPAMI.2018.2889473>.

Kwadwo Daddy Nyame Owusu Boakye

VectrixDB

Lightweight Vector Database with Embedded Machine Learning
Models

mmr_rerank function

<dl><dt>query_vector</dt>
<dd>Query embedding</dd>
<dt>doc_vectors</dt>
<dd>Matrix of document embeddings</dd>
<dt>doc_ids</dt>
<dd>Vector of document IDs</dd>
<dt>scores</dt>
<dd>Initial relevance scores</dd>
<dt>lambda</dt>
<dd>Diversity parameter (0-1)</dd>
<dt>limit</dt>
<dd>Number of results to return</dd></dl>

Arguments

Maximal Marginal Relevance (MMR) reranking — mmr_rerank

<dl>

<dt>query_vector</dt>
<dd>Query embedding</dd>


<dt>doc_vectors</dt>
<dd>Matrix of document embeddings</dd>


<dt>doc_ids</dt>
<dd>Vector of document IDs</dd>


<dt>scores</dt>
<dd>Initial relevance scores</dd>


<dt>lambda</dt>
<dd>Diversity parameter (0-1)</dd>


<dt>limit</dt>
<dd>Number of results to return</dd>

</dl>

mmr_rerank: Maximal Marginal Relevance (MMR) reranking

Description

Usage

Value

Arguments