Knowledge Extraction Technology and Application for Text Data in Military Field

The extraction of intelligence knowledge from unstructured military texts plays an important role in intelligence early warning and command decision-making. The text data in the military field is small and the field is highly correlated. The learning-based method does not have sufficient accuracy and coverage in the extraction of field knowledge. Aiming at this problem, a knowledge extraction method of unstructured intelligence texts combining rules and deep learning methods is proposed. The rules describing the relationship between concepts are generated according to the characteristics of military intelligence texts. The rules are used to add constraints to the deep learning process and the rule model is regarded as the basic model for deep learning. This method can not only complete intelligence knowledge extraction more highly, but also can effectively reduce the cost of manual labeling. In addition, a military intelligence knowledge graph in military field is built based on this, it can be effectively applied to the correlation analysis problems between intelligence.