A comparison of ontology-based and reference-set-based semantic annotation frameworks
Quratulain Rajput, Sajjad Ali Haider · Procedia Computer Science · 2011
A lot of information on World Wide Web, specially the one present on online ad portals, is available in unstructured and ungrammatical format. This unstructured nature of data makes the searching of relevant information a difficult task. The difficulty can be overcome if the contents are semantically annotated. The paper compares two recently proposed semantic annotation frameworks, namely BNOSA and Phoebus that aim to annotate data available on online ad portals. BNOSA performs semantic annotation by utilizing ontology for knowledge building, rule generation and information extraction and Bayesian networks for conflict resolution and missing value prediction. Phoebus, on the other hand, utilizes a collection of known entities, called reference set, and support vector machines to perform semantic annotation. The paper applies both frameworks on same data sets and analyzes their strengths and weaknesses.