Emergent ontology discovered from folksonomies
Feng Chung Wu · Memorial University Research Repository (Memorial University) · 2015
Collaborative tagging websites or systems allow users to associate freely-determined keywords (tags) with a particular resource. The collection of users’ tags and resources is referred to as a folksonomy. Unlike traditional forms of metadata, the meaning of, and relationships between, tags are not rigorously defined, limiting the usefulness of tag-based metadata. We propose a novel approach to enrich tagging systems by constructing a tag ontology that captures semantic relationships among tags. We first consider regularities that can be exploited in a folksonomy. Then, we show how user-level tag vocabulary can be used for tag meaning disambiguation. Following this, we introduce a distance model to calculate the relatedness of two sets of resources within a folksonomy, and use this to develop a method for discovering tag relations. A series of experiments we conducted demonstrate the effectiveness of the method. We conclude the thesis with example use cases where our method can be applied to improve folksonomy data organization and queries. [illustration]