Research and Application on Domain Ontology Learning Method Based on LDA

Considering the problem of multi-source heterogeneous cross-media text information in the field of aviation safety is difficult to share, the paper proposed a domain ontology learning method for civil aviation emergency management. The use of adaptive the NLPIR word segmentation and filtering methods to obtain the candidate term dataset. LDA topic model of domain ontology was designed, through the LDA model training of Gibbs sampling and topic inference to realize the related terms of domain ontology concept core extraction. The construction method of basic semantic relation recognition rules was studied based on the LDA topic probability distribution. The recognition and implementation of the concept and its related term basic semantic relations were presented. Experimental results show that the method can effectively solves the problem of automatic updating of concepts and relations in large-scale domain ontology, and it provided a good data support for sharing and reasoning of civil aviation emergency cross-media information under the environment of big data.