US 12,481,686 B1
Systems and methods for improved data curation and integration
Neema Kechettira Uthappa, Phoenix, AZ (US); Jennifer L. Gilbert, New York, NY (US); Steven F. Reed, New York, NY (US); A. R. Nandakumar Reddy, New York, NY (US); Subhajit Basu, New York, NY (US); Srinivasa Rao Maddili, New York, NY (US); Kunal Chakrabarty, New York, NY (US); and Mujen Inanc, New York, NY (US)
Assigned to MCKINSEY & COMPANY, INC., New York, NY (US)
Filed by MCKINSEY & COMPANY, INC., New York, NY (US)
Filed on Nov. 20, 2024, as Appl. No. 18/954,057.
Int. Cl. G06F 16/33 (2025.01); G06F 16/332 (2019.01); G06F 16/3329 (2025.01); G06F 40/30 (2020.01); H04L 51/02 (2022.01); G06F 16/334 (2025.01); G06F 40/211 (2020.01); G10L 15/18 (2013.01)
CPC G06F 16/3329 (2019.01) [G06F 16/3325 (2019.01); G06F 40/30 (2020.01); H04L 51/02 (2013.01); G06F 16/3347 (2019.01); G06F 40/211 (2020.01); G10L 15/1815 (2013.01); G10L 15/1822 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
receiving, by one or more processors, an initial query from a user as part of a chat session;
querying, by the one or more processors, an external database to obtain a set of keywords, a set of topics, and a set of other queries that are pertinent to the initial query;
embedding, by the one or more processors, articles from a knowledge database and articles from a curated database into a database of embedded articles;
selecting, by the one or more processors, a set of embedded articles from the database of embedded articles, wherein each article in the set of embedded articles has a corresponding similarity value surpassing a threshold value when compared to the initial query using a text similarity metric;
appending, by the one or more processors, (i) the set of keywords, the set of topics, and the set of other queries to the initial query to obtain an external data query and (ii) the set of embedded articles to the initial query to obtain an internal data query;
generating, by the one or more processors, (i) an external output by applying a large language model (LLM) to the external data query and (ii) an internal output by applying the LLM to the internal data query;
combining, by the one or more processors, the internal output with the external output to obtain a combined output;
executing, by the one or more processors, an artificial intelligence (AI) agent configured to generate one or more prompts for the LLM using a chain of thoughts approach and multi-hop query generation that cause the LLM to (i) summarize and rank content provided in the combined output and (ii) generate one or more intermediate reasoning steps indicating how the LLM reached the combined output; and
causing, by the one or more processors, the combined output to be displayed to a user during the chat session.