Knowledge Organization (KO) has historically been used to structure biological knowledge, from taxonomy to ontologies. This becomes increasingly challenging as life sciences evolve into a data-intensive domain. The advent of artificial intelligence (AI) has enabled knowledge organization systems (KOSs) to assume active roles in computational workflows rather than serve as passive repositories. This thematic review examines the evolution of KOSs in AI-augmented biological research by situating them within scientific paradigmatic and epistemological shifts. By synthesizing foundational theories from library and information science, philosophy of science, and biological systematics, we propose the Knowledge Organization Analysis Framework (KOAF) to capture bio-KOSs’ developments across functional sophistication, automation degree in system construction, and reasoning and inference capability. Representative empirical studies show that bio-KOSs enable semantic interoperability and data integration, while also contributing to hypothesis generation and reasoning. We argue that advanced bio-KOSs increasingly function as epistemic agents in scientific discovery. This transformation marks KOSs as theoretical frameworks shaping scientific inquiry through AI-KO convergence and highlights the need for future research on accountability, epistemic integrity, and scientific trustworthiness in AI-driven knowledge discovery.

This work introduces OntoBio, a biographical ontology grounded in Linked Data principles. The study reviews existing ontologies to identify limitations in their ability to represent the diversity and complexity of individual lives. To address these gaps, we developed OntoBio using a tripartite framework encompassing Personality, Environment and Milieu, and Achievements/Milestones. This framework is supported by an expressive vocabulary designed to capture the multifaceted nature of biographical knowledge. OntoBio provides a structural foundation for the semantic integration of existing ontologies that only partially represent biographical data. We present a set of use-case scenarios that motivated OntoBio’s development and highlight its key modeling features. To ensure robustness and interoperability, OntoBio was developed following the Yet Another Methodology for large-scale faceted Ontology Construction (YAMO) methodology and in compliance with World Wide Web Consortium (W3C) standards, including Resource Description Framework (RDF), Web Ontology Language (OWL), and SPARQL Protocol and RDF Query Language (SPARQL). Finally, we evaluate the ontology by constructing a biographical knowledge graph illustrating the life of a prominent individual.

Editorial records of a knowledge organization system (KOS) are useful for tracking the provenance of how and why a concept evolves over time. In this study, we explore the use of Large Language Models (LLMs) in obtaining structured provenance information from editorial records of KOSs. Specifically, this study focuses on one type of provenance, namely, “Warrant”, which refers to external sources for decision-making on KOS changes. This study presents examples based on four Dewey Decimal Classification (DDC) editorial exhibits, each exhibit containing a discussion and proposed actions for change on a DDC topic. To explore whether LLMs can be used to extract Warrant from these DDC exhibits, we design experiments to test the models' consistency and factual accuracy. We use GPT-4o-mini with the Retrieval Augmented Generation (RAG) approach. For the system and user instructions, chain-of-thought and few-shot prompting strategies are used. For consistency, we test the impact of repeated prompting on ChatGPT's performance; for factual accuracy, we assess whether varying temperature settings yield divergent outputs. Our findings show that consistency and factual accuracy are maintained on most categories of warrant information (Document, Literature, and Concept Scheme), with an average F1-score greater than 70%. For extracting “Concept”, the performance is low, with an average F1-score ranging from 30–40%. This demonstrates that ChatGPT is promising for extracting most warrant information from editorial records but still requires a human-in-the-loop verification step to fact-check the concepts extracted. Finally, a recommended process for KOS provenance documentation using LLMs is provided.

Objective: It aims to map, analyze thematically, semantically, and discursively the articles published in the area of Knowledge Organization within the five published volumes of the events of the International Society of Knowledge Organization Brazil (ISKO-Brasil) and the eight volumes of the North American Symposium on Knowledge Organization (NASKO). Methodology: The methodology used was to survey pre-established pivot statements of the scientific production published in the annals of both events with the help of Sketch Engine software as a tool and approach 6 – Historical studies of structures and services of information in domains, 8 – Epistemological and critical studies of different paradigms, assumptions and interests in domains, 10 – Studies of structures and institutions of scientific and professional communication in a domain, from Hjørland (2017), and 13 – Discourse analysis of domain analysis as a methodological contribution, from Smiraglia (2015) and Barros (2023). Results: The survey yielded 131 articles for ISKO-Brasil and 132 documents for NASKO. It was found that, even though the two corpora analyzed are within ISKO's scope, there are divergences regarding the understanding of concepts, as well as their relationship with the epistemological discussion of the area and convergences concerning the concepts of 1) Domain Analysis; 2) Organization Systems; 3) Concept Theory; 4) Classification Systems. Conclusion: The analysis made it possible to envision Knowledge Organization as a theoretical and applied area based on Concept Theory and Domain Analysis.

Controlled vocabularies are widely used in Knowl­edge Organization Systems (KOS); however; they are criticized for perpetuating biases, being slow to change, and not reflecting the language of many groups and cultures. This paper examines and challenges these criticisms by identifying five forms of vocabulary control in use today (minimal control, passive control, post hoc control, flexible control, and rigid control), and studying their effect on the subject indexing and subject tagging processes from the perspective of novice indexers and taggers. The study uses a mixed methods approach, including a survey, a think-aloud protocol that was employed while participants indexed and tagged documents, and a retrospective interview, to better understand participants’ actions, thoughts, and reactions during the indexing and tagging process. The study explores how controlled vocabularies make indexers and taggers feel controlled, in control, or out of control while they index and tag; measures how different forms of vocabulary control affect coextensiveness between aboutness statements and indexing terms; and analyzes how these experiences inform the criticisms of controlled vocabularies, particularly in relation to how vocabulary control aids and hinders racial and social justice in indexes. The study finds that the form of vocabulary control has a significant effect on the subject indexing and subject tagging processes, including an effect on participants’ feelings of control during the processes, an effect on how participants transformed concepts during subject representation, and an effect on how participants navigate complex issues related to racial and social justice.