Research On Ontology Evolution In Open Environments And Its Application In Information Extraction

Ontologies form the core of Semantic Web systems. It is well recognized within Sematic Web that ontologies and Information Extraction (IE) are combined in a cyclic knowledge processing: ontologies are used for interpreting the text at the right context for IE to be efficient and accurate, while IE extracts new knowledge from the text, to be integrated in the ontology. Various efforts have been devoted to the research of different aspects of ontology evolution and IE in recent years. However, ontologies tend to change and evolve over time, ontology evolution is a time consuming and error-prone task, and relies on considerable input and decisions from a user with knowledge representation skill, especially in open network environments. For example, problems arises from the adaptation of IE systems to varying distribution of information items, axioms learning in ontology construction, ontology changes and their sideeffects, and the computation of semantic differences between ontology versions, etc. In this paper, we investigate approaches to these problems in open network environments with the state of the art in ontology envolution and IE.Based on the fact that an ontology can be changed concurrently in open enviroments, this paper proposes a basic frame for handling ontology changes. At first we point out all possible relation types between any two ontology change sequences including dependence relation, directly conflict relation, indirectly conflict relation and compossible relation. According to directly conflict and indirectly conflict, the course of ontology evolution is parted into three phases. The the first phase answers for searching all sets of ontology change sequences which are in directly conflict or indirectly conflict from the prime ontology change sequence set. In the second phase, the prime ontology change sequence set is parted into a group of maximum and sequential ontology change sequence sets according to conflict set. Ontology change sequences in each maximum and sequential ontology change sequence set are executed as their dependence relation. At last, two real cases are given to show usefulness of our approach.In order to improve running efficiency of these algorithms for detecting differences among ontology versions, a novel approach based on concept lattice are proposed. Ontology versions are mapped to formal objects and these taxonomic relationships of ontology versions are mapped to formal attributes. And then the ontology versions space are expressed as version lattice. Based on version lattice, a set of specific algorithms for difference detection with parameter and without parameter are proposed. We not only proved our algorithms but also show high running efficiency by analyzing these algorithms.To deal with disorders among information items, information item missing and information items with multi values in the field of IE, we propose a hierarchical extraction model based on ontology for extracting information on the Web. To do so, we first parse a web document into an extended DOM tree, and then map location and other characteristics of an information item to the corresponding parameters of hierarchical extraction model. At last the initial hierarchical Hidden Markov Model is obtained by using induction algorithm and then it is revised by ontology in order to improve its expression ability. Experiments show that the algorithm has higher precise ratio.At last, we use JAVA program language to implement a prototype software system for ontology evolution and Web information extraction in open network enviroments.