Artificial intelligence custom fields

1 . A computer-implemented method, comprising:

receiving, from a graphical user interface (GUI), a custom rule comprising a query request for information in an electronic document stored in a data repository;

generating, using a processor, a vectorized rule from the custom rule;

selecting, using the processor, a subset of vectorized portions from a set of vectorized portions of the electronic document stored in a vector database corresponding to the vectorized rule;

sending, using the processor, the vectorized rule and the subset of vectorized portions to a machine learning (ML) model, wherein the machine learning model includes at least one of: an artificial neural network (ANN), a generative artificial intelligence (AI) model, a large language model (LLM), a transformer model, or any combination thereof; and

receiving, using the processor, a query response to the query request from the ML model, the query response comprising a summary of textual information contained within one or more vectorized portions from the subset of vectorized portions, or the query response comprising a comment on the textual information contained within the one or more vectorized portions from the subset of vectorized portions, and providing the query response to the GUI for display.

2 . The method of claim 1 , comprising:

determining, using the processor, that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

sending, using the processor, the vectorized rule and an image of the electronic document to a visual ML model; and

receiving, using the processor, a query response to the query request from the visual ML model, the query response comprising a summary of textual information from an image document portion of the image of the electronic document corresponding to the vectorized rule.

3 . The method of claim 1 , comprising:

determining, using the processor, that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

sending, using the processor, the vectorized rule and an image of the electronic document to a visual ML model;

receiving, using the processor, an indicator that the image of the electronic document does not contain any image document portions with textual information that corresponds to the vectorized rule from the visual ML model; and

generating, using the processor, a query response to the query request, the query response comprising a comment indicating that the electronic document does not contain any information corresponding to the vectorized rule.

4 . The method of claim 1 , comprising presenting, using the processor, a name for the custom rule, the query request from the custom rule, and the query response to the query request on the GUI.

5 . The method of claim 1 , comprising:

generating, using the processor, a similarity score for a candidate matching pair comprising a vectorized portion from the set of vectorized portions and the vectorized rule;

ranking, using the processor, the candidate matching pairs based on the similarity scores to form a set of ranked candidate matching pairs; and

selecting, using the processor, the subset of vectorized portions from the set of ranked candidate matching pairs.

6 . The method of claim 1 , comprising:

receiving, using the processor, a second subset of vectorized portions from the subset of vectorized portions identified by the ML model;

sending, using the processor, the vectorized rule and the second subset of vectorized portions to the ML model; and

receiving, using the processor, the query response to the query request from the ML model, the query response comprising a summary of information contained within a single vectorized portion from the second subset of vectorized portions identified by the ML model, or the query response comprising a comment on the information contained within the single vectorized portion.

7 . The method of claim 1 , further comprising:

generating, using the processor, the comment based on at least one vectorized portion, the comment indicating a presence or absence of the information requested by the query request of the custom rule; and

presenting, using the processor, a name of the custom rule, the query request of the custom rule, and the comment to the query request in the GUI.

8 . A computing apparatus, comprising:

at least one processor; and

a memory storing instructions that, when executed by the at least one processor, causes the at least one processor to:

receive a custom rule comprising a query request for information in an electronic document stored in a data repository from a graphical user interface (GUI);

generate a vectorized rule from the custom rule;

select a subset of vectorized portions from a set of vectorized portions of the electronic document stored in a vector database corresponding to the vectorized rule;

send the vectorized rule and the subset of vectorized portions to a machine learning (ML) model, wherein the machine learning model includes at least one of: an artificial neural network (ANN), a generative artificial intelligence (AI) model, a large language model (LLM), a transformer model, or any combination thereof; and

receive a query response to the query request from the ML model, the query response comprising a summary of textual information contained within one or more vectorized portions from the subset of vectorized portions, or the query response comprising a comment on the textual information contained within the one or more vectorized portions from the subset of vectorized portions, and provide the query response to the GUI for display.

9 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

determine that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

send the vectorized rule and an image of the electronic document to a visual ML model; and

receive a query response to the query request from the visual ML model, the query response comprising a summary of textual information from an image document portion of the image of the electronic document corresponding to the vectorized rule.

10 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

determine that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

send the vectorized rule and an image of the electronic document to a visual ML model;

receive an indicator that the image of the electronic document does not contain any image document portions with textual information that corresponds to the vectorized rule from the visual ML model; and

generate the query response comprising a comment indicating that the electronic document does not contain any information corresponding to the vectorized rule.

11 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to present a name for the custom rule, the query request for the custom rule, and the query response to the query request on the GUI.

12 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

generate a similarity score for a candidate matching pair comprising a vectorized portion from the set of vectorized portions and the vectorized rule;

rank the candidate matching pairs based on the similarity scores to form a set of ranked candidate matching pairs; and

select the subset of vectorized portions from the set of ranked candidate matching pairs.

13 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

receive a second subset of vectorized portions from the subset of vectorized portions identified by the ML model;

send the vectorized rule and the second subset of vectorized portions to the ML model; and

receive the query response to the query request from the ML model, the query response comprising a summary of information contained within a single vectorized portion from the second subset of vectorized portions identified by the ML model, or the query response comprising a comment on the information contained within the single vectorized portion.

14 . The computing apparatus of claim 8 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

generate the comment based on at least one vectorized portion, the comment indicating a presence or absence of the information requested by the query request of the custom rule; and

present a name for the custom rule, the query request of the custom rule, and the comment in the GUI.

15 . A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by at least one processor, cause the at least one processor to:

receive, from a graphical user interface (GUI), a custom rule comprising a query request for information in an electronic document stored in a data repository;

generate a vectorized rule from the custom rule;

select a subset of vectorized portions from a set of vectorized portions of the electronic document stored in a vector database corresponding to the vectorized rule;

send the vectorized rule and the subset of vectorized portions to a machine learning (ML) model, wherein the machine learning model includes at least one of: an artificial neural network (ANN), a generative artificial intelligence (AI) model, a large language model (LLM), a transformer model, or any combination thereof; and

receive a query response to the query request from the ML model, the query response comprising a summary of textual information contained within one or more vectorized portions from the subset of vectorized portions, or the query response comprising a comment on the textual information contained within the one or more vectorized portions from the subset of vectorized portions, and provide the query response to the GUI for display.

16 . The computer-readable storage medium of claim 15 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

determine that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

send the vectorized rule and an image of the electronic document to a visual ML model; and

receive a query response to the query request from the visual ML model, the query response comprising a summary of textual information from an image document portion of the image of the electronic document corresponding to the vectorized rule.

17 . The computer-readable storage medium of claim 15 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

determine that the information contained within the one or more vectorized portions is not associated with the vectorized rule by the ML model;

send the vectorized rule and an image of the electronic document to a visual ML model;

receive an indicator that the image of the electronic document does not contain any image document portions with textual information that corresponds to the vectorized rule from the visual ML model; and

generate the query response comprising a comment indicating that the electronic document does not contain any information corresponding to the vectorized rule.

18 . The computer-readable storage medium of claim 15 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

generate a similarity score for a candidate matching pair comprising a vectorized portion from the set of vectorized portions and the vectorized rule;

rank the candidate matching pairs based on the similarity scores to form a set of ranked candidate matching pairs; and

select the subset of vectorized portions from the set of ranked candidate matching pairs.

19 . The computer-readable storage medium of claim 15 , comprising instructions that when executed by the at least one processor, cause the at least one processor to:

receive a second subset of vectorized portions from the subset of vectorized portions identified by the ML model;

send the vectorized rule and the second subset of vectorized portions to the ML model; and

receive the query response to the query request from the ML model, the query response comprising a summary of information contained within a single vectorized portion from the second subset of vectorized portions identified by the ML model, or the query response comprising a comment on the information contained within the single vectorized portion.