Citation Parsing and Analysis with Language Models
Parth Sarin, Juan Pablo Alperin
TL;DR
The study addresses global inequities in scholarly indexing by using open-weight language models to annotate citations into an indexable format. It builds a 2,000-pairs dataset from Garnett/PKP and Open Research Europe, aligning plaintext citations with JATS markup via a similarity measure $s(p,c) = 1 - \frac{d(p,c)}{\max\{|p|,|c|\}}$ and threshold $0.75$ to ensure fidelity. Evaluating eleven open-weight LMs, the authors show high accuracy on citation components, often exceeding state-of-the-art baselines, with the smallest model Qwen3-0.6B achieving strong results after $2^5$ passes, indicating post-training can yield compact, robust parsers. The work highlights practical paths for deploying offline, multilingual citation parsing to improve research indexing and metascientific analysis, and outlines future directions like distillation, constrained decoding, RLVR, and offline metadata enrichment to broaden impact.
Abstract
A key type of resource needed to address global inequalities in knowledge production and dissemination is a tool that can support journals in understanding how knowledge circulates. The absence of such a tool has resulted in comparatively less information about networks of knowledge sharing in the Global South. In turn, this gap authorizes the exclusion of researchers and scholars from the South in indexing services, reinforcing colonial arrangements that de-center and minoritize those scholars. In order to support citation network tracking on a global scale, we investigate the capacity of open-weight language models to mark up manuscript citations in an indexable format. We assembled a dataset of matched plaintext and annotated citations from preprints and published research papers. Then, we evaluated a number of open-weight language models on the annotation task. We find that, even out of the box, today's language models achieve high levels of accuracy on identifying the constituent components of each citation, outperforming state-of-the-art methods. Moreover, the smallest model we evaluated, Qwen3-0.6B, can parse all fields with high accuracy in $2^5$ passes, suggesting that post-training is likely to be effective in producing small, robust citation parsing models. Such a tool could greatly improve the fidelity of citation networks and thus meaningfully improve research indexing and discovery, as well as further metascientific research.