Granted Patent B1

US 12,512,216 · App. 18/957,811 · Granted Dec 30, 2025

Inventors: Blake Browder (Dallas, TX); Joy Figarsky (Little Rock, AR)

Assignee: Behavioral Health Operations, LLC

G16H40/20G16H50/20

Abstract

Claims (59)

1 . An apparatus for determining a code as a function of subject data, wherein the apparatus comprises:

at least a computing device, wherein the computing device comprises:

a memory; and

at least a processor communicatively connected to the memory, wherein the memory contains instructions configuring the at least a processor to:

collect, using the at least a processor, subject data;

receive, using the at least a processor, user input associated with the subject data;

determine, using a first encoder, portions of the user input, wherein the portions comprise tokens and noise;

identify, using a second encoder, the tokens of the portions, wherein identifying the tokens comprises:

extracting a context datum from the portions;

comparing segments of the portions to predefined data, wherein comparing the segments of the portions to the predefined data comprises using a cross-encoder by:

 inputting, simultaneously, both the user input and the predefined into the cross-encoder;

 assessing, by the cross-encoder, a degree of match between one or more segments of the portions of the user input and the predefined data by jointly encoding a combined context of the one or more segments and the predefined data based on at least the extracted context datum; and

 producing, by the cross-encoder, a similarity score representative of the degree of match between the one or more segments of the user input and the predefined data;

filtering the noise from the portions as a function of a comparison of the segments to the predefined data based on the similarity score being within a range of values; and

identifying the tokens as a function of a comparison of the segments to the predefined data based on the similarity score not being within the range of values;

assign, using the second encoder, a code to the identified tokens, wherein the second encoder is iteratively trained using system feedback wherein the system feedback generates at least an error message as a function of an unexpected condition, wherein assigning the code comprises:

determining at least a classification for the tokens;

matching, using at least a reference database, the tokens to one or more codes associated with the at least a reference database; and

identifying the code as a function of the subject data; and

display, using a downstream device, the code.

2 . The apparatus of claim 1 , wherein the apparatus is further configured to collect the subject data from electronic subject records.

3 . The apparatus of claim 1 , wherein the first encoder comprises a natural language processor trained using a first dataset comprising historical portions corresponding to historical noise and historical tokens.

4 . The apparatus of claim 1 , wherein the second encoder comprises a large language model trained using a second dataset comprising historical tokens corresponding to historical code.

5 . The apparatus of claim 1 , wherein the second encoder is iteratively trained using system feedback comprising a correction datum based on the user input.

6 . The apparatus of claim 1 , wherein the apparatus further comprises a processing module configured to:

organize the subject data into predefined categories based on a temporal datum; and

filter the subject data based on a frequency datum.

7 . The apparatus of claim 1 , wherein identifying the tokens of the portions comprises detecting data elements that correspond to one or more codes associated with the at least a reference database.

8 . The apparatus of claim 1 , wherein identifying the code as a function of the subject data comprises analyzing context data of the user input, wherein identifying the code comprises comparing the context data of the user input to a key datum of the subject data.

9 . The apparatus of claim 1 , wherein the apparatus is further configured to assign the code based on a hierarchical system.

10 . The apparatus of claim 9 , wherein the hierarchical system is configured to generate a probability score as a function of a confidence datum.

11 . A method for determining a code as a function of subject data, wherein the method comprises:

collecting, using at least a processor, subject data;

receiving, using the at least a processor, user input associated with the subject data;

determining, using a first encoder, portions of the user input, wherein the portions comprise tokens and noise;

identifying, using a second encoder, the tokens of the portions, wherein identifying the tokens comprises:

extracting a context datum from the portions;

comparing segments of the portions to predefined data, wherein comparing the segments of the portions to the predefined data comprises using a cross-encoder by:

inputting, simultaneously, both the user input and the predefined into the cross-encoder;

assessing, by the cross-encoder, a degree of match between one or more segments of the portions of the user input and the predefined data by jointly encoding a combined context of the one or more segments and the predefined data based on at least the extracted context datum; and

producing, by the cross-encoder, a similarity score representative of the degree of match between the one or more segments of the user input and the predefined data;

filtering the noise from the portions as a function of a comparison of the segments to the predefined data based on the similarity score being within a range of values; and

identifying the tokens as a function of a comparison of the segments to the predefined data based on the similarity score not being within the range of values;

assigning, using the second encoder, a code to the tokens, wherein the second encoder is iteratively trained using system feedback wherein the system feedback generates at least an error message as a function of an unexpected condition, wherein assigning the code comprises:

determining at least a classification for the tokens;

matching, using at least a reference database, the tokens to one or more codes associated with the at least a reference database; and

identifying the code as a function of the subject data; and

displaying, using a downstream device, the code.

12 . The method of claim 11 , wherein collecting the subject data comprises collecting the subject data from electronic subject records.

13 . The method of claim 11 , wherein the first encoder comprises a natural language processor trained using a first dataset comprising historical portions corresponding to historical noise and historical tokens.

14 . The method of claim 11 , wherein the second encoder comprises a large language model trained using a second dataset comprising historical tokens corresponding to historical code.

15 . The method of claim 11 , wherein the second encoder is iteratively trained using system feedback comprising a correction datum based on the user input.

16 . The method of claim 11 further comprising:

organizing, using a processing module, the subject data into predefined categories based on a temporal datum; and

filtering, using the processing module, the subject data based on a frequency datum.

17 . The method of claim 11 , wherein identifying the tokens of the portions comprises detecting data elements that correspond to one or more codes associated with the at least a reference database.

18 . The method of claim 11 , wherein identifying the code as a function of the subject data comprises analyzing context data of the user input, wherein identifying the code comprises comparing the context data of the user input to a key datum of the subject data.

19 . The method of claim 11 , wherein assigning the code comprises assigning the code based on a hierarchical system.

20 . The method of claim 19 , wherein the hierarchical system is configured to generate a probability score as a function of a confidence datum.

Assignments (3)

References Cited (16)