Mining a multilingual association dictionary from Wikipedia for cross-language information retrieval
Journal of the American Society for Information Science and Technology
| Hi-index | 0.00 |
In the context of the Semantic Web, many ontology-related operations can be boiled down to one fundamental task: finding as accurately as possible the semantics hiding beneath the superficial representation of ontological entities. This, however, is not an easy task due to the ambiguous nature of semantics and a lack of systematic engineering method to guide how we comprehend semantics. We acknowledge the gap between human cognition and knowledge representation formalisms: even though precise logic formulae can be used as the canonical representation of ontological entities, understanding of such formulae may vary. A feasible solution to juxtaposing semantics interpretation, therefore, is to reflect such cognitive variations. In this paper, we propose an approximation of semantics using sets of words/phrases, referred to as WɪKɪmantic vectors. These vectors are emerged through a set of well-tuned methods gradually surfacing the semantics that remain implicit otherwise. Given a concept, we first identify its conceptual niche amongst its neighbours in the graph representation of the ontology. We generate a natural language paraphrases of the isolated sub-graph and project this textual description upon a large document repository. WɪKɪmantic vectors are then drawn from the document repository. We evaluated each of the aforementioned steps by way of user study.