Several Concept-based Knowledge Representations And Related Approaches

Formal Concept Analysis, Ontology and Event are three emerging concept based knowledge representations. They research on the nature of concepts and the relations between concepts, but the emphases of them are different with each other. Formal Concept Analysis focuses mainly on the formation of the concepts. Ontology more stresses on how to represent the concepts and their relations. And Event emphasizes on verbal concepts to represent the structure of concepts. This paper focuses on these three models to resolve their respective existent problems, and to combine them together according to their different emphases. The content of this paper includes:1. A large collection of formal concepts can be a hedge of the application of fuzzy concept lattice. It is not directly comprehensible for a user. Development of methods which help to overcome the problem of large number of extracted formal concepts is thus an important task. This paper proposes a method for reducing the size of fuzzy concept lattices. The principle is to cluster concepts according to a distance measure between concepts, and that is based on the truth values from a fuzzy formal context. The method proposed is clustering-based reduction algorithm of fuzzy concept lattice by the similarity between concepts measured according to concept vectors. It forms concept hierarchy that is the reduction of fuzzy concept lattice. The experiments show that the time complexity of this reduction is reasonable, and that the compression rate of the concept lattice is important while preserving the accuracy of classification methods based on the fuzzy concept lattices.2. To help handling the issues of automatic acquisition of concepts and their relations for constructing ontology, this paper proposes an ontology construction method based on fuzzy concept hierarchy. With excellent mathematical nature and sound construction algorithms and tools, Formal Concept Analysis is used for mining concepts especially abstract concepts and the generalization and specialization relations between concepts. The concepts and their relations in concept hierarchy are mapped into the concepts and taxonomic relations between them in ontology automatically. Thereby, it reduces the participation of domain experts in the process of ontology construction, and achieves a higher degree of automation.3. Classical Formal Concept Analysis cannot deal with the information expressed by intervals in practice. Interval Formal Concept Analysis is proposed here whose core data structure is interval concept lattice. It incorporates attribute decomposition based interval attribute scaling to extend Formal Concept Analysis to have the capacity of processing interval information. The experimental results show that the construction algorithm has reasonable performance on the complexity.4. An event representation model and the approach of event extraction are produced. This model uses the event tuple to express event. Comparing with the event triple, it is more flexible because the triple is too simplistic to represent the entire event. The event tuple not only considers related named entities (the time, location and participants of the event), and also considers important related nouns and terms which helps to find and represent the abstract and professional events. Based on this event model, the event extraction approach is produced to extract events from texts based on natural language processing. It includes the lexical analysis and syntactic analysis of the texts and extraction of semantic elements and events from texts automatically. This event model extends the existing event models. The method of event extraction enhances existing methods of event extraction. They help to promote the development of the theoretical models and provide the application tools for artificial intelligence, summarization, text processing and other fields.5. The bottleneck of ontology learning is the acquisition of relations between concepts especially non-taxonomy relations. Event-based knowledge acquisition technologies are proposed for ontology learning to overcome this problem. It starts from a small core ontology constructed by domain experts. The concepts and their taxonomic and non-taxonomic relations are learned by event-based knowledge acquisition technologies automatically for the construction of the aimed ontology. Under the instruction of domain experts or ontology engineers, the ontology evaluation makes sure if the ontology is learned properly and decides whether it is necessary to repeat the iterative learning processes. If the decision is the ontology learning process need to take iteratively, the last aimed ontology becomes the core ontology of this time and the domain experts should to supply some new domain corpus as the input corpus for further event-based learning of the concepts and relations. So the process starts again from this new core ontology and this time the domain corpus is different. The experimental results indicate that the concepts and relations in the aim ontology increase. Most of the domain concepts are learned from general ontology and most of relations are acquired from event based concept and relation learning. This method offers an improvement in quality of ontology.6. An Event and Formal Concept Analysis based summarization method is proposed. It based on the event extraction method produced in this paper to extract the events from the documents. Then, the redundant and contradict events are removed from extracted events for the construction of formal context. The concept lattice is built from the formal context to determine the relevance between events for the computation of the significances of events. Together with the context of event, the significance of sentence in the document is measured for the extraction of summarization. This summarization method is evaluated on the standard evaluation corpus. Evaluation results show that it achieves good results of the evaluation. It confirms the effect of the method. This summarization method can be used for enriching the instance of ontology for the improvement the application of ontology.In summary, this paper studies on Formal Concept Analysis, Ontology and Event these three concept based knowledge representations. Each of them exists problems need to solve. In fact, some of these problems can be solved through the combination of themselves. In particular, Formal Concept Analysis and Event can help ontology construction. The combination research of Formal Concept Analysis and Ontology on the theory and applications is an important research area. It is to achieve excellent research results by introducing event into ontology and knowledge management areas. The combination of these three concept based knowledge representations has good prospects and important significance.