Abstract: METHOD AND SYSTEM FOR GENERATING USER ROLE-SPECIFIC RESPONSES THROUGH LARGE LANGUAGE MODELS ABSTRACT This disclosure relates to method (300) and system (200) for generating user role-specific responses (211) through Large Language Models (LLMs). The method (300) includes receiving (303) a user query (210) from a user account. The user account is associated with a user role from a plurality of user roles. The method (300) includes combining (304), based on the user role, the user query (210) with contextual information corresponding to the user query (210) to obtain a combined query. The method (300) includes inputting (305) the combined query to an LLM. The method (300) includes generating (306) a user role-specific response (211) corresponding to the combined query through the LLM. The method (300) includes rendering (307) the user role-specific response (211) to a display of a user device associated with the user account. [To be published with FIG. 2]
1. A method (300) for generating user role-specific responses through Large Language
Models (LLMs), the method (300) comprising:
5 receiving (303), by a computing device (101), a user query (210) from a user
account, wherein the user account is associated with a user role from a plurality of user
roles;
based on the user role, combining (304), by the computing device (101), the
user query (210) with contextual information corresponding to the user query (210) to
10 obtain a combined query;
inputting (305), by the computing device (101), the combined query to an LLM;
generating (306), by the computing device (101), a user role-specific response
(211) corresponding to the combined query through the LLM; and
rendering (307), by the computing device (101), the user role-specific response
15 (211) to a display of a user device associated with the user account.
2. The method as claimed in claim 1, comprising extracting (406), by the computing
device (101), the contextual information from one or more data sources, wherein the
one or more data sources comprise a plurality of vector embeddings corresponding to
20 the plurality of user roles, and wherein the contextual information comprises
information associated with the user role stored in one or more of the plurality of vector
embeddings.
3. The method (300) as claimed in claim 1, further comprising extracting (301), by the
25 computing device (101), the user role from data associated with the user account.
4. The method (300) as claimed in claim 1, further comprising receiving (302), by the
computing device (101), the user role as an input from the user device.
42
5. The method as claimed in claim 1, further comprising assigning, by the computing
device (101), data access control to the user account based on the user role, wherein
the data access control is based on a relationship of the user role with remaining of the
plurality of user roles.
5
6. The method as claimed in claim 1, further comprising:
receiving (401), by the computing device (101), source data (501) comprising
user role data corresponding to the plurality of user roles from one or more data
sources;
10 generating (402), by the computing device (101), one or more data chunks from
the user role data;
transforming (403), by the computing device (101), the one or more data chunks
into a plurality of vector embeddings;
associating (404), by the computing device (101), the plurality of vector
15 embeddings with the one or more data chunks based on the user role using a plurality
of links to obtain a plurality of linked vector embeddings; and
storing (405), by the computing device (101), the plurality of linked vector
embeddings in the one or more data sources in an arrangement based on predefined
relationships between the corresponding plurality of user roles.
20
7. A system (200) for generating user role-specific responses through Large Language
Models (LLMs), the system (200) comprising:
a processing circuitry (201); and
a memory (202) communicatively coupled to the processing circuitry (201),
25 wherein the memory (202) stores instructions, which when executed by the processing
circuitry (201), cause the processing circuitry (201) to:
receive (303) a user query (210) from a user account, wherein the user
account is associated with a user role from a plurality of user roles,
43
combine (304), based on the user role, the user query (210) with
contextual information corresponding to the user query (210) to obtain a
combined query,
input (305) the combined query to an LLM,
5 generate (306) a user role-specific response (211) corresponding to the
combined query through the LLM, and
render (307) the user role-specific response (211) to a display of a user
device associated with the user account.
10 8. The system (200) as claimed in claim 7, wherein the instructions, on execution, cause
the processing circuitry (201) to:
extract (406) the contextual information from one or more data sources,
wherein the one or more data sources comprise a plurality of vector embeddings
corresponding to the plurality of user roles, and wherein the contextual information
15 comprises information associated with the user role stored in one or more of the
plurality of vector embeddings.
9. The system (200) as claimed in claim 7, wherein the instructions, on execution, cause
the processing circuitry (201) to assign data access control to the user account based
20 on the user role, wherein the data access control is based on a relationship of the user
role with remaining of the plurality of user roles.
10. The system (200) as claimed in claim 7, wherein the instructions, on execution,
cause the processing circuitry (201) to:
25 receive (401) source data (501) comprising user role data corresponding to the
plurality of user roles from the one or more data sources;
generate (402) one or more data chunks from the user role data;
transform (403) the one or more data chunks into a plurality of vector
embeddings;
44
associate (404) the plurality of vector embeddings with the one or more data
chunks based on the user role using a plurality of links to obtain a plurality of linked
vector embeddings; and
store (405) the plurality of linked vector embeddings in the one or more data
5 sources in an arrangement based on predefined relationships between the
corresponding plurality of user roles.
| # | Name | Date |
|---|---|---|
| 1 | 202341060054-STATEMENT OF UNDERTAKING (FORM 3) [06-09-2023(online)].pdf | 2023-09-06 |
| 2 | 202341060054-PROVISIONAL SPECIFICATION [06-09-2023(online)].pdf | 2023-09-06 |
| 3 | 202341060054-POWER OF AUTHORITY [06-09-2023(online)].pdf | 2023-09-06 |
| 4 | 202341060054-FORM 1 [06-09-2023(online)].pdf | 2023-09-06 |
| 5 | 202341060054-DRAWINGS [06-09-2023(online)].pdf | 2023-09-06 |
| 6 | 202341060054-DECLARATION OF INVENTORSHIP (FORM 5) [06-09-2023(online)].pdf | 2023-09-06 |
| 7 | 202341060054-FORM 18 [20-02-2024(online)].pdf | 2024-02-20 |
| 8 | 202341060054-DRAWING [20-02-2024(online)].pdf | 2024-02-20 |
| 9 | 202341060054-CORRESPONDENCE-OTHERS [20-02-2024(online)].pdf | 2024-02-20 |
| 10 | 202341060054-COMPLETE SPECIFICATION [20-02-2024(online)].pdf | 2024-02-20 |
| 11 | 202341060054-Power of Attorney [10-01-2025(online)].pdf | 2025-01-10 |
| 12 | 202341060054-Form 1 (Submitted on date of filing) [10-01-2025(online)].pdf | 2025-01-10 |
| 13 | 202341060054-Covering Letter [10-01-2025(online)].pdf | 2025-01-10 |