Risk information review method based on content matching

Abstract

The application discloses a risk information review method based on content matching, relates to the technical field of natural language processing, and comprises the following steps: acquiring a first risk content segment, analyzing and extracting a first risk semantic core unit and an original bearing structure, establishing an archive set after marking a multi-dimensional state enumeration value; backtracking to collect current public texts, shielding text units matched with the first hit keyword, extracting a candidate bearing structure and establishing a slot set; cross-comparing the above sets, generating a risk semantic core migration chain when the keyword overlapping degree of the candidate bearing structure is zero, the structure role membership is compatible with the whitelist, and no exclusion rule is hit; generating a second early warning information object according to the migration chain, and outputting a target positioning coordinate; and the application breaks through the limitation of surface character comparison, restores the topological path of the hidden transfer of the risk intention, and improves the accuracy of content compliance review.

Images