Knowledge acquisition method from domain text based on theme logic model and artificial neural network

In order to acquire knowledge from domain text such as failure analysis text of aviation product, a framework is proposed to enhance the efficiency and accuracy of knowledge acquisition. In this framework, sentence templates are defined to extract the meta-knowledge and RDF is used to describe the extracted knowledge. After the preprocessing steps, the authors propose a new model: theme logic model (TLM) to present all the themes of a piece of text and the logical relations among different themes. In this model, the text of each theme can be represented as an attribute–value vector based on domain ontology. Meanwhile, the logical relations are the domain knowledge to be acquired. The theme logic model then will be transformed to the training set of the artificial neural network to acquire the failure analysis knowledge. After training process, acquired knowledge will be extracted by SD method from the artificial neural network and represented by rules. Therefore, a prototype is developed to acquire knowledge from failure analysis reports of aviation product. Empirical results show that the framework can acquire knowledge from domain text efficiently.