| CPC G09B 7/06 (2013.01) [G06F 40/279 (2020.01); G06F 40/30 (2020.01)] | 18 Claims |
|
1. A method of generating knowledge assessment items directed to a subject area, the method comprising:
receiving, from a user interface, a set of model knowledge assessment items collectively directed to the subject area;
grouping the set of model knowledge assessment items into a plurality of similar item groups using numeric features corresponding to the model knowledge assessment items, wherein the plurality of similar item groups include selected items of the set of model knowledge assessment items that are semantically similar to other model knowledge assessment items in the respective similar item groups;
generating a conditioning input for each of the plurality of similar item groups to by ordering the model knowledge assessment items based on a distance between said model knowledge assessment items and a centroid in each of the plurality of similar item groups;
receiving, from a transformer-based natural language generation (NLG) model, raw assessment items, wherein each of the raw assessment items is generated responsive to providing the conditioning input for one of the plurality of similar item groups to the transformer-based NLG model, and wherein each of the raw assessment items is generated based on a contextual embedding vector for each token in the conditioning input;
receiving, from the user interface, specifications for the knowledge assessment items;
in response to receiving the specifications, identifying the knowledge assessment items from the raw assessment items, wherein identifying the knowledge assessment items from the raw assessment items comprises comparing features of the raw assessment items to the specifications for the knowledge assessment items; and
providing the knowledge assessment items to the user interface, wherein the knowledge assessment items are collectively directed to the subject area.
|