relevant-retrieval-ratio v1.0.3

Relevant Retrieval Ratio

This repository contains the code required to compute the Relevant Retrieval Ratio ($R^3$), an implementation agnostic link prioritization performance metric that compares prioritization algorithms to an optimal oracle algorithm.

Installation

To install the necessary dependencies, run the following command:

    yarn install

Addtionally, this package requires the JDK.

Usage

The metric's computation is handled in the class RunLinkTraversalPerformanceMetrics. The method RunMetricAll computes the $R^{3}$ value and requires the following inputs:

• Edgelist: Represents the queried topology. Expects edges as arrays of numbers in the following format [start, end, weight]

• Relevant Document Sets: The number representation of sets of documents needed to produce each result.

• Traversal Path: The path taken by the engine during traversal. Expects edges as arrays of numbers in the following format start, end, weight

• Seed Documents: The initial number representations of documents from which the traversal starts.

To run the solver, the input edgelist, root documents, and query-relevant documents are written to a file in .stp format located in heuristic-solver/input/full-topology/. The solver then reads this file and returns the heuristically optimal traversal path.