Artificial intelligence (AI)-based regulatory data processing system

A method of processing domain-specific regulatory documents comprising:

• accessing a domain-specific regulatory text corpus that includes prior domain-specific documents having regulatory information pertaining to a specific domain;
• training a topic extraction model to extract one or more domain-specific topics via unsupervised learning on topic training data that includes the prior domain-specific documents in the regulatory text corpus wherein the prior domain-specific documents include domain-specific topics;
• training an entity identification model to identify domain-specific entities via supervised learning on entity training data within the prior domain-specific documents in the regulatory text corpus wherein identifying the domain-specific entities includes identifying linguistic features in the prior domain-specific documents;
• training a section classification model via supervised learning on section training data to classify portions of regulatory text into one of a plurality of predetermined sections wherein classifying the portions into one of the plurality of predetermined sections includes identifying linguistic features in the prior domain-specific documents;
• identifying one or more of the topics and one or more of the domain-specific entities from a received domain-specific document including regulatory text using the trained topic extraction model and the trained entity identification model;
• classifying one or more portions of the received domain-specific document into one of the plurality of predetermined sections using the trained section classification model;
• automatically identifying one or more actions to be executed to implement regulations in the received domain-specific document based on verbs in the topics identified in the received domain-specific document;
• and
• notifying one or more users regarding the actions to be executed for implementing the regulations in the received domain-specific document.

A non-transitory processor-readable storage medium comprising machine-readable instructions that cause a processor to:

• receive a domain-specific document for analysis, the received domain-specific document including regulatory text associated with a specific domain;
• extract topics in the received domain-specific document using a topic extraction model, the topic extraction model being trained via unsupervised training on prior domain-specific documents in a regulatory text corpus for identifying the topics from the received domain-specific document;
• identify and classify one or more entities in the received domain-specific document as belonging to entities by an entity identification model, the entity identification model being trained via supervised learning on the prior domain-specific documents;
• classify one or more portions in the received domain-specific document as belonging to one of a plurality of predetermined sections by a section identification model, the section identification model being trained via supervised learning on the prior domain-specific documents;
• identify one or more actions to be executed for implementing processes by identifying the one or more sections corresponding to requirements sections and the topics that include verbs extracted from the received domain-specific document;
• obtain a corresponding priority for each of the one or more actions based at least on date entities determined from the one or more entities;
• and
• generate one or more notifications regarding the one or more actions and the corresponding priorities for each of the one or more actions.