A pipeline for matching bibliographic references with incomplete metadata: experiments with Crossref and OpenCitations

TL;DR

This paper presents a methodology for matching bibliographic references with incomplete metadata by analyzing unstructured text and partial data, improving citation link creation between entities in bibliographic databases.

Contribution

It introduces a heuristic-based tool and methodology for mapping references with incomplete metadata to existing entities, enhancing citation network completeness.

Findings

High matching precision achieved

Effective integration of partial metadata

Recall limitations indicate need for further improvements

Abstract

While Crossref makes available more than 1.8 billion bibliographic references from publications for which it provides a DOI, more than 698 million of these references do not specify a DOI, making the creation of a formal citation link from the citing entity and the cited entity problematic. In this article, we propose an analysis of Crossref bibliographic references to show how we can use the unstructured text defining such references and the available (and partial) metadata specified in them to (a) map them to existing entities included in OpenCitations Meta and, then, (b) to enable the potential inclusion of additional and valid citations link among these entities. We have defined a precise methodology to address the analysis and run it against a manually defined Gold Standard and a subset of Crossref. While the heuristic-based tool developed has demonstrated strong matching precision and effective metadata integration, its recall limitations highlight the necessity of further enhancements to address metadata inconsistencies and leverage additional sources of citation data.