Conceptual Clustering of Documents for Automatic Ontology Generation

摘要

In Information retrieval, Keyword based retrieval is unsatisfactory for user needs since it can't always retrieve relevant words according to the concept. Since different words can represent the same concept (polysemy) and one word can represent different concepts (homonymy), mapping problem will lead to word sense Disambiguation. Through the implementation of domain dependent ontology, concept based information retrieval (IR) can be achieved. Since Semantic concept extraction from keywords is the initial phase for automatic construction of ontology process, this paper propose an effective method for it. Reuters 21578 is used as the input of this process, followed by indexing, training and clustering using self-Organizing Map. Based on the feature vector, the clustering of documents are formed using automatic concept selections, in order to make the hierarchy. Clusters are represented hierarchically based on the topics assigned .Ontology will be generated automatically for each cluster, based on the topic assigned.