Jack Kausch

How do we design classification systems that embrace ambiguity and are highly multilingual? This dissertation presents two speculative knowledge organization systems called ‘Horapollo Ontology’ and ‘Real Character Language Ontology’. These systems use icons to represent cross-linguistic semantic primitives. The systems ask two research questions. The first is: how do we determine which categories are most well attested cross-linguistically? The second is: how do we represent categories with easily interpretable images? Taking a documentalist approach to the role of knowledge organization, these questions are situated within the context of Francis Bacon’s definition of “Real Character” from the history of knowledge organization, and present a problematic to challenge the unspoken assumptions of universality in the field of library and information science, particularly in the use of authority control to manage ambiguity. Bacon’s ideas about universal language are further situated in the context of the Renaissance Hieroglyphic tradition, a tradition deriving from the encyclopedia of Horapollo, from which the Horapollo Ontology gains its name. Combining research creation, a mixed methods quantitative content analysis, information visualization, and the construction of an information retrieval system, this project is grounded in contemporary linguistic linked open data (LLOD) standards. The Cross-linguistic linked data (CLLD) Concepticon dataset is used as a source of crosslinguistic concepts derived from a systematic review of linguistic literature, while the Noun Project API is used to collect crowd-curated images associated these concepts. The concept sets are coded to determine which concepts are best represented as cross-lingual images. Following this, research creation is used to visualize and dramatize the results of the research. The 26 best performing concepts are visualized as a font, Horapollo 1.3. The same images from the font are defined at unique URIs in the Horapollo ontology. 14 of these concepts are reused to form the Real Character Language Ontology, an experimental ontology which uses fuzzy logic to operate over a multilingual vector database for an information retrieval system using Wikipedia articles. The result is a knowledge organization system which is a work of art intended to provoke questions concerning our assumptions about the cross-linguistic universality of knowledge, and also demonstrate expanded potentials for information retrieval in the era of machine learning.