Jenny Copara | Searcharxiv

arXiv2025

MeSH Concept Relevance and Knowledge Evolution: A Data-driven Perspective

The Medical Subject Headings (MeSH), one of the main knowledge organization systems in the biomedical domain, continuously evolves to reflect the latest scientific discoveries in health and life sciences. Previous research has focused on quantifying information in MeSH primarily through its hierarchical structure. In this work, we propose a data-driven approach based on information theory and network analysis to quantify the relevance of MeSH concepts. Our method leverages article annotations and their citation networks to compute four aspects of relevance -- informativeness, usefulness, disruptiveness, and influence -- over time. Using both the citation network and the MeSH hierarchy, we compute these relevance aspects and apply an aggregation algorithm to propagate scores to parent nodes. We evaluated our approach on MeSH terminology changes and showed that it effectively captures the evolution of concepts. The mean relevance of evolving concepts is higher compared to concepts that remained unchanged ($2.09E-03$ vs. $8.46E-04$). Moreover, we validated the framework by analyzing retracted articles and found that concepts used to annotate retracted articles (mean relevance: 0.17) differ substantially from those annotating non-retracted ones (mean relevance: 0.15). Overall, the proposed framework provides an effective method for ranking concept relevance and can support the maintenance of evolving knowledge organization systems.

cs.SI↗

arXiv2020

Named entity recognition in chemical patents using ensemble of contextual language models

Chemical patent documents describe a broad range of applications holding key reaction and compound information, such as chemical structure, reaction formulas, and molecular properties. These informational entities should be first identified in text passages to be utilized in downstream tasks. Text mining provides means to extract relevant information from chemical patents through information extraction techniques. As part of the Information Extraction task of the Cheminformatics Elsevier Melbourne University challenge, in this work we study the effectiveness of contextualized language models to extract reaction information in chemical patents. We assess transformer architectures trained on a generic and specialised corpora to propose a new ensemble model. Our best model, based on a majority ensemble approach, achieves an exact F1-score of 92.30% and a relaxed F1-score of 96.24%. The results show that ensemble of contextualized language models can provide an effective method to extract information from chemical patents.

cs.CL↗