The invention discloses a formal semantic reasoning and
deep learning-based
natural language knowledge mining method. The method comprises the following steps: a
machine learning step used for combining a formal semantic reasoning method with a
machine learning method and performing
machine learning including semanteme, an intention learning step used for learning a descriptive intention of a to-be-mined
natural language, a text topic extraction step used for analyzing a text topic content and
paragraph content descriptive intention according to an LDA model, and a document structure classification
model building step used for performing model training according to a deep
convolutional neural network, automatically modifying and perfecting the neural network in a training process and building an automatic document structure classifier. According to the method, a non-structured data group represented by literatures, voices and the like is directly processed and the intention is reasoned by depending on the combination of automated semantic reasoning, a
deep learning technology and
natural language processing, thereby obtaining a real intention of text description; and mined and retrieved knowledge is used for analysis and construction of a new
knowledge structure, so that
knowledge mining and retrieval can be performed in an efficient, intelligent, learnable and self-evolution manner.