CPC G06Q 10/0635 (2013.01) [G06F 16/3329 (2019.01); G06F 16/353 (2019.01); G06N 5/022 (2013.01); G06Q 10/06375 (2013.01); G06Q 50/188 (2013.01)] | 20 Claims |
1. A method of determining a risk category for a contract clause, the method comprising:
receiving, at a server, seeds of text from contract documents, each seed having been assigned a risk category from a set of risk categories, the set of risk categories including an evergreen warranty, unlimited liability, unfavorable terms, missing protection, transferred liability, indemnity failure, and no baseline period;
passing the seeds of text through a multi-head attention mechanism and a feed forward mechanism to produce processed seeds of text;
feeding an artificial intelligence specificity model training inputs including:
the processed seeds of text,
the assigned risk category for each seed of text, and
a contract ontology defining pertinent terms and relationships between various aspects of the contract,
thereby training the specificity model to classify contract clauses into a risk category;
receiving, at the server and from a first computing device, a first contract document and a first query regarding a risk associated with a first clause in the first contract document;
feeding the first clause into the specificity model;
automatically determining, via the specificity model, that the first clause falls under a first risk category of the set of risk categories;
extracting, from the first contract document, a penalty clause associated with the first risk category;
causing a computing device to present a message identifying the first risk category for the first clause as well as information about the penalty clause;
receiving feedback via the computing device regarding an accuracy of the first risk category classification of the first clause; and
automatically retraining the specificity model based on the received feedback.
|