KNOWLEDGE ORGANIZATION IN CAPTURING EMOTIONAL SEMANTICS AND ORAL HISTORY CONSTRUCTS
Traversing the boundaries of language: Using knowledge organisation to investigate the semantics of emotion across English and Sinhala
By Malithi Alahapperuma
Emotions are a critical part of the human experience. Despite this, putting feelings into words is never easy. Some emotions are so complex, it is difficult to find the right words to convey the feeling. Simultaneously, some emotion terms are so nuanced, it is hard to completely fathom their meaning. This challenge exacerbates when communicating emotions across languages, as emotions rarely have objective anchor points. As a result, cross-linguistic emotion expression may become an exercise of merely translating words. However, in the semantics of emotion, linguistic equivalence rarely guarantees conceptual equivalence. Emotion terms often embody the social and cultural norms of a linguistic community, which are hard to convey with just one word or phrase.
This talk presents research published in the proceedings of the 18th Metadata and Semantics Research Conference, exploring a systematic inquiry into using knowledge organisation systems to better understand and encode emotionally weighted material. The study focuses on English and Sinhala, a language whose emotion lexicon is deeply shaped by its cultural and religious heritage, as a case study for examining how emotion concepts traverse across linguistic boundaries. The study maps emotion vocabulary across the two languages using a combination of synonym rings and discourse analysis. The work approaches the problem through three pillars of inquiry, cognition, linguistics, and socio-cultural factors, to explore the ways in which individuals perceive and express emotions across English and Sinhala. The goal of such an inquiry is to develop a more grounded framework for understanding the semantics of emotions in cross-linguistic settings.
Ontology Development with Large Language Models for Oral History Network Construction and Hermeneutical Analysis
By Andreas Vlachidis
Oral history is situated, intersubjective, and performative, and not merely a record of recalled events but a testimony of how individuals construct meaning from the past. The talk presents work with a corpus of 25 interviews conducted between 2012 and 2015 that comprise narratives of formation, disruption, and change in the history of Digital Humanities. Migrating oral history testimonies into digital environments risks an epistemological tension where formal knowledge representations can strip away the very nuances and contextuality that make oral testimony analytically significant. Addressing this challenge, this presentation introduces the Hermeneutic Ontology Engineering methodology developed within the MeDoraH (Mixed-Methods Digital Oral History) project. We explore how Large Language Models can be employed to accelerate entity detection and relationship identification in the process of conceptual modelling, while critically assessing their limitations in ensuring logical consistency, terminological correctness, and ontological rigour. We present a hybrid methodology that harnesses LLMs' capabilities within structured workflows that integrate critical human expertise at key decision points. Our experience demonstrates that LLMs offer substantial generative potential yet require principled human intervention and iterative refinement cycles to maintain the semantic precision essential for robust knowledge representation. The presentation offers insights into whether LLMs can reliably support formal ontology engineering for humanities corpora, what kinds of human intervention mechanisms are necessary, and how human expertise can be optimally integrated within these processes.
