Automated extraction of deontological statements through a multilevel analysis of legal acts

Abstract

Currently, large amounts of data are available in text form. This makes the automatic semantic analysis of natural language texts a topical task. This very complex task can be rather effectively solved in case when consideration is limited by certain specific kind of texts. The paper describes the developed technology of extraction of deontological statements from legal acts through a multilevel analysis of text documents combining linguistic and semantic methods. The technology is tested on the example of Russian legislative documents and shows its potential effectiveness. Experiments have shown that certain types of Universal Dependencies relations with high accuracy identify the components of deontological statements.