| CPC G06Q 10/0833 (2013.01) [G06Q 10/0832 (2013.01); G06Q 10/0838 (2013.01); G06T 7/0002 (2013.01)] | 20 Claims |
|
1. A computer program product comprising:
a program of instructions tangibly embodied on a non-transitory computer readable medium wherein, when the instructions are executed on a processor, the processor causes operations to be performed to automatically inspect contents of closed packages, the operations comprising:
apply, in response to receiving a signal from a customs broker package inspection station corresponding to text that indicates contents of a closed package, the text being external to the package, a natural language processing model to generate a content description profile, wherein the signal further indicates an originator of the package and a destination jurisdiction;
apply a content characterization model to the content description profile to generate a content attributes profile associating the contents of the package with at least one predetermined content attribute; and,
generate a jurisdiction entry compliance indicator (JECI) by a second set of operations comprising:
select predetermined permissions rules as a function of the destination jurisdiction and the content attributes profile;
assign predetermined permission attributes as a function of the originator of the package and the content attributes profile;
apply the predetermined permissions rules to the predetermined permission attributes to generate the JECI that indicates a permissibility of the contents of the package to enter the destination jurisdiction;
apply at least one predetermined confidence criterion to the JECI; and,
when the JECI meets the at least one predetermined confidence criterion, then automatically generate and transmit an indication to customs broker personnel operating the inspection station not to open the package and perform manual inspection of the contents,
dynamically training the content characterization model based on historical content attributes associated with packages corresponding to a JECI that did not meet the at least one predetermined confidence criterion, the training comprising:
train an initial content characterization model with training data to generate a trained content characterization model, the training data comprising predetermined content description data and content attribute data associated with the predetermined content description data;
apply the trained content characterization model to testing data to generate a training metric as a function of a comparison between predicted content attributes and model content attributes associated with the testing data; and,
when the training metric does not meet a predetermined training criterion, then repeat at least the steps of train and apply,
wherein the at least one predetermined confidence criterion corresponds to accuracy of the content attribute characterizing the contents of the closed package and to conformance of the contents of the closed package to the predetermined permissions rules and the predetermined permissions attributes.
|