Research Of Ontology Learning Methods Based On Formal Concept Analysis

Ontology is used to describe or express a set of concepts or terms for particular field, it can be used to organize the abstract of knowledge and can also be used to describe the specific areas of knowledge. Ontology which was considered as a model of expressing the construct of concept hierarchy and semantics has been applied widely in the field of computer science. Therefore how to build ontology effective and useful has become a hot problem for researching. Ontology learning is the method and technology to build ontology automatically or semi-automatically. Based on ontology structure, the mission of ontology learning includes the concepts acquisition, the relationship between concepts acquisition and axiom acquisition. And currently there are some research to the acquisition of concepts and their relationship, but the research to axiom acquisition is seldom. Most of the ontology building is a particular object-oriented, so it still does not have a standard to use. And because of the domain ontology's specificity, it still needed experts through the process of the ontology building. Around this background, the thesis did the research and comparison to the existing methods and then introduced the Formal Concept Analysis method to the study of ontology learning and finally put forwarded a text-oriented ontology building model based on Formal Concept Analysis (TOOBM).TOOBM model includes the whole process from domain text to the generation of ontology, including:1. Text pre-processingThe domain text needs to be pre-processing, excavated the concepts in the text. First of all, the original text needs to be converted to the text with part of speech. Secondly, the thesis designed a concepts generate algorithm. The algorithm firstly removes the redundancy in the text, and then converts the long sentence to a simple sentence, and at last matches the objects and properties to form the concepts. This part finishes the excavating of the domain concepts.2. Formal Concept AnalysisFormal Concept Analysis is a branch of applied mathematics. The keystone of the ontology building based on Formal Concept Analysis is to help find out the relationship between the concepts. The formal concepts are ambiguity in the documents, so how to convert them to the relationship between the concepts is the difficult problem. And the Formal Concept Analysis can solve the problem well. The set of concepts which was pre-processing in the first part could generate concept lattice through the Formal Concept Analysis. This part can acquire the relationship between the concepts.3. Ontology prototype generationIn this part, the thesis uses the reduction method to reduce the concept lattice. It can generate redundant after Formal Concept Analysis, such as the redundant of the properties, so it must reduce the concept lattice. It is necessary to ensure each concept has an unique name. so some of the concepts need to be renamed. And because of the particularity of the domain knowledge, some concepts'name need to be given by domain experts. Through the above steps, the result of this part is the reduction lattice based on ontology expression, that is, the ontology class hierarchy.4. OWL ontology generationWeb Ontology Language, OWL is the standard which W3C2004 recommended, its aim is providing more language to support for the expression of the semantic and reasoning. The process of the ontology building based on OWL is finished by Protégé. First of all, ontology class is generated based on the hierarchy, then disjoint the related classes. And it created the properties by the verbs which stored in the first part.In this process, it needs to set up the property information, such as domain and range and so on. Then created the restriction information and finally link all of the related classes through the properties. Therefore this part completes the process of the ontology building based on OWL ontology description language.The thesis realizes the generation of the domain OWL ontology through a food drink text by TOOBM. It excavated the domain concepts from the original food and drink text after the text pre-processing. Then through the revision and confirmation by experts, it forms the set of concepts, that is, the formal background. And then it acquired the concept of hierarchy by the analyzing of Formal Concept Analysis, that is, the concept lattice. Further more, the concept lattice needs to be reduced and renamed to generate the ontology class hierarchy. And at last the OWL domain ontology is generated based on the class hierarchy through Protégé. Further, the thesis tests the TOOBM model by using data sets. The data sets come from a food science web set, it includes 31 domain texts. Through the the participation of experts in the field, the thesis realized the domain OWL ontology building by TOOBM model and realized the semi-automatic construction of ontology.In summary, the thesis puts forward a text-oriented ontology building model based on Formal Concept Analysis, TOOBM. The model generates the OWL ontology effectively and usefully through the domain texts. But in the entire design process, it still can not do without the participation of domain experts, so the model is the semi-automatic construction of ontology. In the future the author will have to improve on the above work, including: to optimize the concepts generate algorithm and concept lattice construction algorithm, to improve the operating efficiency and to reduce the participation of experts and other manual work and finally to complete the text-oriented automatic ontology building based on Formal Concept Analysis.