| This thesis presents a technique to automatically generate ranked lists of relevant terms for each concept in a concept hierarchy (ODP). The terms are obtained from the concept hierarchy itself, and from WordNet. Each concept is mapped to a WordNet synset using the concept title, whenever possible. When there are multiple synsets to choose from, where each synset corresponds to a different meaning of the concept title, the correct synset is identified using a new automatic sense-disambiguation algorithm, developed as part of this thesis. Terms obtained from the concept hierarchy are ranked for each concept, on the basis of their frequencies of occurrence within that concept, and the total number of concepts they occur in. Terms obtained from WordNet are ranked on the basis of the semantic strength. These two lists of terms are combined, and their ranking is based on their statistical as well as semantic strengths. |
