This research proposes an unsupervised ontology based feature matching approach to address the semantic gap problem in information retrieval. The approach involves natural language processing, semantic feature extraction and selection using a light weight user-oriented ontology. The approach comprises four stages: (1) user-oriented ontology building, (2) semantic feature extraction for building vectors representing information objects, (3) semantic feature matching using the user-oriented ontology, and (4) measuring the similarity between the information objects. The evaluation conducted shows that the ontology based approach consistently outperforms the term-based retrieval approach.
|Title of host publication||Frontiers in Artificial Intelligence and Applications|
|Subtitle of host publication||Advances in Knowledge-Based and Intelligent Information and Engineering Systems|
|Publication status||Published - 2013|