Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Text Processing

Research on Domain Ontology Concept Extraction and Relation Extraction

Author YangFen
Tutor ZhangYuFang
School Chongqing University
Course Computer System Architecture
Keywords Ontology Learning Context Concept Extraction Relation extraction Semantic relevancy
CLC TP391.1
Type Master's thesis
Year 2010
Downloads 249
Quotes 2
Download Dissertation

With the development of society, the increasingly strong demand for digital information, IT faces many challenges, such as the representation of information, knowledge sharing, software reuse. How vast amount of information on the network organization, management, maintenance and reuse, and provide prompt and effective service for enterprise users has become a widespread concern in the field of scientific research. The body as a shared conceptual model, have been more and more people are concerned about and have been widely used in many fields of artificial intelligence, knowledge engineering, semantic retrieval. The manual build ontology is a tedious and hard task, need to spend a lot of time and resources, especially in building domain ontology also need the participation of experts in the field, is a major bottleneck in the development of the body. To solve the problem, people began to try to build ontology automatic or semi-automatic ontology learning. Ontology learning is the use of statistics, machine learning, natural language processing technology, semi-automatic or automatically from the existing text, HTML documents, databases and other data sources to obtain the desired ontology technology, the current study focuses on the concept as well as the relationship between the extraction. The traditional ontology learning method based on statistical methods, and to focus more on the concept, the relationship between the ontology semantic polymerizable ignored semantic extraction results for the relationship between the concept and the concept of extraction accuracy lacking. To solve the above problem, this paper proposes a filtering mechanism based ontology learning method, using the the vocabulary context constructor concept vector space model, cosine similarity calculated to represent the semantic correlation between vocabulary. Using semantically related degree of extraction of the concept, the concept of the relationship between the purpose of the filter to increase the accuracy. In addition, on the basis of the concept of relational learning, the acquisition method of classification relationship explored: the term encompasses a method for classification relationship and set a confidence measure formula to obtain the classification relationship gives credibility metrics. In order to verify the validity of the model, the completion of an ontology learning system, the traditional ontology learning model based on semantic filtering ontology learning comparative experiment. In the evaluation of the experimental results, the paper introduce Hownet lexical similarity calculation software to build the reference standard body to enhance the objectivity of the evaluation of the experimental results by calculating the semantic similarity between words. Experimental results show that the improved model can effectively improve the concept, the precision of the relationship, confirmed the effectiveness of the filter based on semantic ontology learning model.

Related Dissertations
More Dissertations