Label-based information deficiency processing
There is a need for more effective and efficient information deficiency processing. This need can be addressed by, for example, solutions for performing/executing label-based information deficiency processing. In one example, a method includes receiving a predictive input associated with a predictive entity based on the predictive input; determining a plurality of encoding probability values for the predictive entity based on the predictive input; determining a plurality of attention-based encoding vectors for the predictive entity based on the plurality of encoding probability values; determining an encoding deficiency prediction for the predictive entity, wherein the encoding deficiency prediction indicates a deficient subset of a plurality of encoding designations; and for each encoding designation in the deficient subset, performing a corresponding prediction-based action.
1. A computer-implemented method comprising:
receiving, by one or more processors, a predictive input associated with a predictive entity;
determining, by the one or more processors, based at least in part on the predictive input and using a first machine learning model comprising a multi-channel convolutional neural network model characterized by a plurality of convolutional channels, a plurality of encoding probability values for the predictive input, wherein each encoding probability value of the plurality of encoding probability values indicates an estimated encoding association of the predictive input with a corresponding encoding designation of a plurality of encoding designations;
determining, by the one or more processors, based at least in part on the predictive input and using a second machine learning model, a plurality of attention-based encoding vectors for the predictive input, wherein each attention-based encoding vector of the plurality of attention-based encoding vectors relate a corresponding encoding designation of the plurality of encoding designations to a related subset of the predictive input;
determining, by the one or more processors and based at least in part on (i) the plurality of encoding probability values, (ii) predictive entity data associated with the predictive entity, and (iii) deficiency utility data associated with the plurality of encoding designations, an encoding deficiency prediction for the predictive input, wherein (a) the encoding deficiency prediction indicates a deficient subset of the plurality of encoding designations and a deficiency utility value for one or more encoding designations in the deficient subset, and (b) the deficiency utility data indicate one or more of an estimated utility cost or an estimated utility reward associated with a deficiency of one or more encoding designations of the plurality of encoding designations; and
for the one or more encoding designations in the deficient subset, initiating, by the one or more processors, the performance of a corresponding prediction-based action based at least in part on the deficiency utility value and a corresponding attention-based encoding vector for the one or more encoding designations.
2. The computer-implemented method of claim 1 , wherein determining the encoding deficiency prediction comprises:
determining, based at least in part on the plurality of encoding probability values and the predictive entity data, the deficient subset;
determining, based at least in part on the deficiency utility data, one or more deficiency utility values for the one or more encoding designations in the deficient subset; and
determining the encoding deficiency prediction based at least in part on the deficient subset and the one or more deficiency utility values for the one or more encoding designations in the deficient subset.
3. The computer-implemented method of claim 2 , wherein determining the deficient subset comprises:
determining, based at least in part on the plurality of encoding probability values, a detected subset of the plurality of encoding designations;
determining, based at least in part on the predictive entity data, a prior subset of the plurality of encoding designations;
determining, based at least in part on the detected subset and the prior subset, a missing subset of the plurality of encoding designations; and
determining the deficient subset based at least in part on the missing subset.
4. The computer-implemented method of claim 3 , wherein determining the deficient subset further comprises:
determining, for one or more encoding designations in the missing subset, one or more designation deficiency likelihood values based at least in part on deficiency likelihood data associated with the plurality of encoding designations; and
determining the deficient subset based at least in part on the one or more designation deficiency likelihood values for the one or more encoding designations in the missing subset.
5. The computer-implemented method of claim 1 , wherein
the first machine learning model is configured to process the predictive input, explanatory metadata associated with the predictive input, and provider metadata associated with the predictive input to generate the plurality of encoding probability values.
6. The computer-implemented method of claim 5 , wherein:
the plurality of convolutional channels include a static channel and a dynamic channel;
the static channel comprises a first vector generated based at least in part on the predictive input that is kept static during training of the first machine learning model; and
the dynamic channel comprises a second vector generated based at least in part on the predictive input that changes during the training of the first machine learning model.
7. The computer-implemented method of claim 1 , wherein determining an attention-based encoding vector of the plurality of attention-based encoding vectors comprises:
generating, based at least in part on the predictive input, explanatory metadata associated with the predictive input, provider metadata associated with the predictive input, and the second machine learning model, a corresponding attention-based discrete distribution over the predictive input for one or more encoding designations of the plurality of encoding designations; and
for the one or more encoding designations of the plurality of encoding designations, determining the attention-based encoding vector based at least in part on the corresponding attention-based discrete distribution associated with the one or more encoding designations.
8. The computer-implemented method of claim 7 , wherein the second machine learning model comprises an attention-based convolutional model that applies a corresponding attention-based mechanism for the one or more encoding designations of the plurality of encoding designations.
9. The computer-implemented method of claim 1 , wherein:
the predictive input is associated with a medical chart;
the predictive entity is associated with a patient profile;
the predictive entity data associated with the patient profile comprise patient documentation data associated with the patient profile;
one or more encoding designations of the plurality of encoding designations are associated with a diagnosis code of a plurality of diagnosis codes; and
the deficient subset is determined based at least in part on one or more diagnosis codes of the plurality of diagnosis codes that are extracted from the medical chart but that are not indicated by the patient documentation data.
10. The computer-implemented method of claim 1 , wherein performing the corresponding prediction-based action for the one or more encoding designations in the deficient subset comprises:
determining, based at least in part on encoding subject matter data for the encoding designation, a corresponding encoding subject matter for the one or more encoding designations;
selecting a responsive agent profile for the one or more encoding designations based at least in part on the corresponding subject matter for the one or more encoding designations;
generating one or more agent instructions for the responsive agent profile based at least in part on the attention-based encoding vector for the one or more encoding designations; and
providing the one or more agent instructions associated with the one or more encoding designations to the corresponding encoding subject matter.
11. The computer-implemented method of claim 10 , wherein generating the one or more agent instructions comprises:
mapping an attention-based encoding vector for the one or more encoding designations to one or more natural language outputs for the one or more encoding designations;
generating an agent instruction interface that comprises the one or more natural language outputs; and
presenting the agent instruction interface to the responsive agent profile.
12. An apparatus comprising one or more processors and at least one memory including program code, the at least one memory and the program code configured to, with the one or more processors, cause the apparatus to at least:
receive a predictive input associated with a predictive entity;
determine, based at least in part on the predictive input and using a first machine learning model comprising a multi-channel convolutional neural network model characterized by a plurality of convolutional channels, a plurality of encoding probability values for the predictive entity, wherein each encoding probability value of the plurality of encoding probability values indicates an estimated encoding association of the predictive input with a corresponding encoding designation of a plurality of encoding designations;
determine, based at least in part on the predictive input and using a second machine learning model, a plurality of attention-based encoding vectors, wherein each attention-based encoding vector of the plurality of attention-based encoding vectors relates a corresponding encoding designation of the plurality of encoding designations to a related subset of the predictive input;
determine, based at least in part on (i) the plurality of encoding probability values, (ii) predictive entity data associated with the predictive entity, and (iii) deficiency utility data associated with the plurality of encoding designations, an encoding deficiency prediction for the predictive input, wherein (a) the encoding deficiency prediction indicates a deficient subset of the plurality of encoding designations and a deficiency utility value for one or more encoding designations in the deficient subset, and (b) the deficiency utility data indicate one or more of an estimated utility cost or an estimated utility reward associated with a deficiency of one or more encoding designations of the plurality of encoding designations; and
for the one or more encoding designations in the deficient subset, initiate the performance of a corresponding prediction-based action based at least in part on the deficiency utility value and a corresponding attention-based encoding vector for the one or more encoding designations.
13. The apparatus of claim 12 , wherein determining the encoding deficiency prediction comprises:
determining, based at least in part on the plurality of encoding probability values and the predictive entity data, the deficient subset;
determining, based at least in part on the deficiency utility data, one or more deficiency utility values for the one or more encoding designations in the deficient subset; and
determining the encoding deficiency prediction based at least in part on the deficient subset and the one or more deficiency utility values for the one or more encoding designations in the deficient subset.
14. The apparatus of claim 13 , wherein determining the deficient subset comprises:
determining, based at least in part on the plurality of encoding probability values, a detected subset of the plurality of encoding designations;
determining, based at least in part on the predictive entity data, a prior subset of the plurality of encoding designations;
determining, based at least in part on the detected subset and the prior subset, a missing subset of the plurality of encoding designations; and
determining the deficient subset based at least in part on the missing subset.
15. The apparatus of claim 12 , wherein
the first machine learning model is configured to process the predictive input, explanatory metadata associated with the predictive input, and provider metadata associated with the predictive input to generate the plurality of encoding probability values.
16. The apparatus of claim 12 , wherein determining an attention-based encoding vector of the plurality of attention-based encoding vectors comprises:
generating, based at least in part on the predictive input, explanatory metadata associated with the predictive input, provider metadata associated with the predictive input, and the second machine learning model, a corresponding attention-based discrete distribution over the predictive input for one or more encoding designations of the plurality of encoding designations; and
for the one or more encoding designations of the plurality of encoding designations, determining the attention-based encoding vector based at least in part on the corresponding attention-based discrete distribution associated with the one or more encoding designations.
17. The apparatus of claim 12 , wherein:
the predictive input is associated with a medical chart;
the predictive entity is associated with a patient profile;
the predictive entity data associated with the patient profile comprise patient documentation data associated with the patient profile;
one or more encoding designations of the plurality of encoding designations are associated with a diagnosis code of a plurality of diagnosis codes; and
the deficient subset is determined based at least in part on one or more diagnosis codes of the plurality of diagnosis codes that are extracted from the medical chart but that are not indicated by the patient documentation data.
18. The apparatus of claim 12 , wherein performing the corresponding prediction-based action for the one or more encoding designations in the deficient subset comprises:
determining, based at least in part on encoding subject matter data for the one or more encoding designations, a corresponding encoding subject matter for the one or more encoding designations;
selecting a responsive agent profile for the one or more encoding designations based at least in part on the corresponding subject matter for the one or more encoding designations;
generating one or more agent instructions for the responsive agent profile based at least in part on the attention-based encoding vector for the one or more encoding designations; and
providing the one or more agent instructions associated with the one or more encoding designations to the corresponding encoding subject matter.
19. The apparatus of claim 18 , wherein generating the one or more agent instructions comprises:
mapping an attention-based encoding vector for the one or more encoding designations to one or more natural language outputs for the one or more encoding designations;
generating an agent instruction interface that comprises the one or more natural language outputs; and
presenting the agent instruction interface to the responsive agent profile.
20. At least one non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions configured, when executed by one or more processors, to:
receive a predictive input associated with a predictive entity;
determine, based at least in part on the predictive input and using a first machine learning model comprising a multi-channel convolutional neural network model characterized by a plurality of convolutional channels, a plurality of encoding probability values for the predictive entity, wherein each encoding probability value of the plurality of encoding probability values indicates an estimated encoding association of the predictive input with a corresponding encoding designation of a plurality of encoding designations;
determine, based at least in part on the predictive input and using a second machine learning model, a plurality of attention-based encoding vectors, wherein each attention-based encoding vector of the plurality of attention-based encoding vectors relates a corresponding encoding designation of the plurality of encoding designations to a related subset of the predictive input;
determine, based at least in part on (i) the plurality of encoding probability values, (ii) predictive entity data associated with the predictive entity, and (iii) deficiency utility data associated with the plurality of encoding designations, an encoding deficiency prediction for the predictive input, wherein (a) the encoding deficiency prediction indicates a deficient subset of the plurality of encoding designations and a deficiency utility value for one or more encoding designations in the deficient subset, and (b) the deficiency utility data indicate one or more of an estimated utility cost or an estimated utility reward associated with a deficiency of one or more encoding designations of the plurality of encoding designations; and
for the one or more encoding designations in the deficient subset, initiate the performance of corresponding prediction-based action based at least in part on the deficiency utility value and a corresponding attention-based encoding vector for the one or more encoding designations.