Granted Patent B1

US 12,253,973 · App. 18/811,016 · Granted Mar 18, 2025

Inventors: Shailesh Gavankar (Princeton, NJ); Afrid Mondal (Karnataka, IN); Keon Park (Yonkers, NY); Sanket Jain (Karnataka, IN); Abhijit Naik (Maharashtra, IN)

Assignee: Morgan Stanley Services Group Inc.

G06F16/152

Abstract

Claims (85)

1. A computer-implemented method for information retrieval via artificial intelligence, the method comprising:

accessing, by at least one computing device configured by executing instructions stored on non-transitory processor readable media, content stored in documents in at least one database;

identifying, by the at least one computing device, contextual chunks of individual ones of the documents, wherein the contextual chunks include portions of content stored in the individual ones of the documents;

generating, by the at least one computing device, embeddings associated with the contextual chunks;

storing, by the at least one computing device, the embeddings in a vector database;

identifying, by the at least one computing device as a function of natural language processing, a plurality of relationships among at least some of the contextual chunks;

generating, by the at least one computing device, relational embeddings using the plurality of identified relationships;

storing, by the at least one computing device, the relational embeddings in a database;

receiving, by the at least one computing device from at least one other computing device, a query for information associated with at least some of the content;

generating, by the at least one computing device, an embedding representing the query;

transmitting, by the at least one computing device to at least one large language model, the embedding representing the query;

receiving, by the at least one computing device from the at least one large language model in response to the embedding representing the query, a response including at least some of the contextual chunks; and

transmitting, by the at least one computing device, the response to a computing device associated with the query.

2. The method of claim 1 , further comprising:

calculating, by the at least one computing device, relevant weights for each of the documents and the contextual chunks.

3. The method of claim 2 , further comprising:

applying, by the at least one computing device, at least some of the relevant weights to at least some of the documents and associated contextual chunks stored in the at least one database.

4. The method of claim 2 , further comprising:

receiving, by the at least one computing device, information associated with a user who provided the query;

wherein the relevant weights are calculated as a function of the information associated with the user.

5. The method of claim 4 , wherein the information associated with the user includes:

authorization of the user to access;

clickstream data associated with the user; and

metadata associated with at least one of the documents.

6. The method of claim 4 , further comprising:

ranking, by the at least one computing device, associated chunks comprised in the response; and

determining, by the at least one computing device as a function of the ranking, whether the user is entitled to receive the response.

7. The method of claim 6 , further comprising:

providing, to the computing device associated with the query, the response if the user is entitled to receive the response.

8. A computer-implemented system for information retrieval via artificial intelligence, the system comprising:

at least one computing device that executes instructions stored on non-transitory processor readable media, wherein the at least one computing device is configured by executing the instructions to:

access content stored in documents in at least one database;

identify contextual chunks of individual ones of the documents, wherein the contextual chunks include portions of content stored in the individual ones of the documents;

generate embeddings associated with the contextual chunks;

store the embeddings in a vector database;

identify, as a function of natural language processing, a plurality of relationships among at least some of the contextual chunks;

generate relational embeddings using the plurality of identified relationships;

store the relational embeddings in a database;

receive, from at least one other computing device, a query for information associated with at least some of the content;

generate an embedding representing the query;

transmit, to at least one large language model, the embedding representing the query;

receive, from the at least one large language model in response to the embedding representing the query, a response including at least some of the contextual chunks; and

transmit, the response to a computing device associated with the query.

9. The system of claim 8 , wherein the at least one computing device is further configured by executing the instructions to:

calculate relevant weights for each of the documents and the contextual chunks.

10. The system of claim 9 , wherein the at least one computing device is further configured by executing the instructions to:

apply at least some of the relevant weights to at least some of the documents and associated contextual chunks stored in the at least one database.

11. The system of claim 9 , wherein the at least one computing device is further configured by executing the instructions to:

receive information associated with a user who provided the query, wherein the relevant weights are calculated as a function of the information associated with the user.

12. The system of claim 11 , wherein the information associated with the user includes:

authorization of the user to access;

clickstream data associated with the user; and

metadata associated with at least one of the documents.

13. The system of claim 11 , wherein the at least one computing device is further configured by executing the instructions to:

rank associated chunks comprised in the response; and

determine, as a function of the ranking, whether the user is entitled to receive the response.

14. The system of claim 13 , wherein the at least one computing device is further configured by executing the instructions to:

provide, to the computing device associated with the query, the response if the user is entitled to receive the response.

15. A non-transitory computer readable medium having computer instructions stored therein that, when executed by at least one computing device, cause the at least one computing device to carry out an automated process, the process comprising:

accessing, by the at least one computing device, content stored in documents in at least one database;

identifying, by the at least one computing device, contextual chunks of individual ones of the documents, wherein the contextual chunks include portions of content stored in the individual ones of the documents;

generating, by the at least one computing device, embeddings associated with the contextual chunks;

storing, by the at least one computing device, the embeddings in a vector database;

identifying, by the at least one computing device as a function of natural language processing, a plurality of relationships among at least some of the contextual chunks;

generating, by the at least one computing device, relational embeddings using the plurality of identified relationships;

storing, by the at least one computing device, the relational embeddings in a database;

receiving, by the at least one computing device from at least one other computing device, a query for information associated with at least some of the content;

generating, by the at least one computing device, an embedding representing the query;

transmitting, by the at least one computing device to at least one large language model, the embedding representing the query;

receiving, by the at least one computing device from the at least one large language model in response to the embedding representing the query, a response including at least some of the contextual chunks; and

transmitting, by the at least one computing device, the response to a computing device associated with the query.

16. The computer readable medium of claim 15 , wherein the process further comprises:

calculating, by the at least one computing device, relevant weights for each of the documents and the contextual chunks.

17. The computer readable medium of claim 16 , wherein the process further comprises:

applying, by the at least one computing device, at least some of the relevant weights to at least some of the documents and associated contextual chunks stored in the at least one database.

18. The computer readable medium of claim 16 , wherein the process further comprises:

receiving, by the at least one computing device, information associated with a user who provided the query;

wherein the relevant weights are calculated as a function of the information associated with the user.

19. The computer readable medium of claim 18 , wherein the information associated with the user includes:

authorization of the user to access;

clickstream data associated with the user; and

metadata associated with at least one of the documents.

20. The computer readable medium of claim 18 , wherein the process further comprises:

ranking, by the at least one computing device, associated chunks comprised in the response; and

determining, by the at least one computing device as a function of the ranking, whether the user is entitled to receive the response.

Assignments (1)

References Cited (10)

Cited By (2)

US 12,450,274 US 12,585,709