Measuring the Accuracy of Linked Data Indices

TL;DR

This paper discusses and compares methods for measuring the accuracy of Linked Data indices, addressing issues of redundancy, outdated information, and the impact on search and caching performance.

Contribution

It introduces and evaluates various measures for index accuracy, analyzing their theoretical and practical effectiveness on real-world Linked Data.

Findings

Different accuracy measures have distinct advantages and disadvantages.

Empirical analysis reveals practical behavior of these measures on real data.

Some measures better capture the impact of outdated data on index quality.

Abstract

Being based on Web technologies, Linked Data is distributed and decentralised in its nature. Hence, for the purpose of finding relevant Linked Data on the Web, search indices play an important role. Also for avoiding network communication overhead and latency, applications rely on indices or caches over Linked Data. These indices and caches are based on local copies of the original data and, thereby, introduce redundancy. Furthermore, as changes at the original Linked Data sources are not automatically propagated to the local copies, there is a risk of having inaccurate indices and caches due to outdated information. In this paper I discuss and compare methods for measuring the accuracy of indices. I will present different measures which have been used in related work and evaluate their advantages and disadvantages from a theoretic point of view as well as from a practical point of view by analysing their behaviour on real world data in an empirical experiment.