Integrated Ontology Learner: Towards Generic Semantic Annotation Framework

To achieve goal of Semantic Web, existing Web documents have to be tagged with semantic information using ontology as a formal conceptualization of a particular domain. However, the manual ontology creation is tedious, expensive, biased, and complex task which can easily result in a knowledge acquisition bottleneck. This paper presents generic and automatic ontology learning approach that uses data from both unstructured and semi structured sources. The approach relies on statistical and neural network techniques (word2vec) to transform implicit knowledge in unstructured text into explicit machine-processable domain knowledge with feature of adaptability into other domains and languages. We experiment the feasibility of the proposed approach for tourism domain using Amharic news collected from Walta news agency and Amharic Wikipedia dump. The experimental result exhibits 78.75% of precision in candidate term extraction, 79.59% of precision in taxonomy induction and 55.00% of precision in specific semantic relation extraction for a morphologically complex Amharic language with limited size corpus.