Knowledge graph construction and application for improving search recalls

Patent metadata

Field	Value
Publication number	US-2026030243-A1
Application number	US-202418787333-A
Country	US
Kind code	A1
Filing date	Jul 29, 2024
Priority date	Jul 29, 2024
Publication date	Jan 29, 2026
Grant date	—

Abstract

Systems and methods of knowledge graph construction are provided. A communication platform accesses communication data on the communication platform associated with an enterprise. The communication platform extracts a plurality of keywords from the communication data. The communication platform creates a knowledge graph comprising a plurality of nodes. The plurality of nodes comprises a plurality of keyword nodes. The communication platform identifies, using a graph random walk algorithm, one or more keyword nodes in the knowledge graph related to a first keyword node to link the first keyword node with the one or more keyword nodes. The communication platform determines a first keyword node embedding for the first keyword node by aggregating information associated with at least the first keyword node and the one or more keyword nodes using a graph convolutional network algorithm. The communication platform attaches the first keyword node embedding to the first keyword node in the knowledge graph.

First claim

1 . A method comprising:
accessing communication data on a communication platform associated with an enterprise; extracting a plurality of keywords from the communication data; creating a knowledge graph comprising a plurality of nodes, the plurality of nodes comprising a plurality of keyword nodes; identifying, using a graph random walk algorithm, one or more keyword nodes in the knowledge graph related to a first keyword node; linking the first keyword node with the one or more keyword nodes; determining a first keyword node embedding for the first keyword node based at least on information associated with the first keyword node and the one or more keyword nodes using a graph convolutional network algorithm; and attaching the first keyword node embedding to the first keyword node in the knowledge graph.

2 . The method of claim 1 , further comprising:
receiving a user search query; determining a query embedding based on the user search query; identifying from the knowledge graph a group of keyword nodes related to user query by comparing the query embedding to a plurality of keyword node embeddings associated with the plurality of keyword nodes in the knowledge graph; retrieving a set of communication data related to the group of keyword nodes; and providing an answer to the user search query based on the set of communication data.

3 . The method of claim 2 , further comprising:
determining a similarity score for a keyword node by comparing the query embedding and the keyword node embedding corresponding to the keyword node in the knowledge graph; and in response to determining the similarity score is greater than a threshold value, retrieving a subset of communication data associated with the keyword node.

4 . The method of claim 1 , further comprising:
updating the knowledge graph based on updates in communication data associated with the enterprise periodically.

5 . The method of claim 1 , wherein the plurality of nodes further comprises a plurality of user nodes corresponding to a plurality of user accounts associated with the enterprise on the communication platform and a plurality of entity nodes corresponding to a plurality of functionalities on the communication platform.

6 . The method of claim 5 , wherein the knowledge graph is a weighted graph, wherein links between the plurality of user nodes and the plurality of entity nodes are weighted based on organization data associated with the enterprise and the communication data.

7 . The method of claim 6 , wherein identifying, using a graph random walk algorithm, one or more keyword node in the knowledge graph related to a first keyword node to link the first keyword node with the one or more keyword nodes comprises:
creating a walk path to a second keyword node within a predetermined degree of separation from the first keyword node in the weighted graph based on linkage weights of links associated with user nodes and entity nodes between the first keyword node and the second keyword node.

8 . A system comprising:
a communications interface; a non-transitory computer-readable medium; and one or more processors communicatively coupled to the communications interface and the non-transitory computer-readable medium, the one or more processors configured to execute processor-executable instructions stored in the non-transitory computer-readable medium to: access communication data on a communication platform associated with an enterprise; extract a plurality of keywords from the communication data; create a knowledge graph comprising a plurality of nodes, the plurality of nodes comprising a plurality of keyword nodes; identify, using a graph random walk algorithm, one or more keyword nodes in the knowledge graph related to a first keyword node; link the first keyword node and the one or more keyword nodes; determine a first keyword node embedding for the first keyword node based at least on information associated with the first keyword node and the one or more keyword nodes using a graph convolutional network algorithm; and attach the first keyword node embedding to the first keyword node in the knowledge graph.

9 . The system of claim 8 , wherein the one or more processors are configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
receive a user search query; determine a query embedding based on the user search query; identify from the knowledge graph a group of keyword nodes related to user query by comparing the query embedding to a plurality of keyword node embeddings associated with the plurality of keyword nodes in the knowledge graph; retrieve a set of communication data related to the group of keyword nodes; and provide an answer to the user search query based on the set of communication data.

10 . The system of claim 9 , wherein the one or more processors are configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
determine a similarity score for a keyword node by comparing the query embedding and the keyword node embedding corresponding to the keyword node in the knowledge graph; and in response to determining the similarity score is greater than a threshold value, retrieve a subset of communication data associated with the keyword node.

11 . The system of claim 8 , wherein the one or more processors are configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
update the knowledge graph based on updates in communication data associated with the enterprise periodically.

12 . The system of claim 9 , wherein the plurality of nodes further comprises a plurality of user nodes corresponding to a plurality of user accounts associated with the enterprise on the communication platform and a plurality of entity nodes corresponding to a plurality of functionalities on the communication platform.

13 . The system of claim 12 , wherein the knowledge graph is a weighted graph, wherein links between the plurality of user nodes and the plurality of entity nodes are weighted based on social relation data and interaction data associated with the plurality of user nodes and the plurality of entity nodes.

14 . The system of claim 13 , wherein the one or more processors are configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
create a walk path to a second keyword node within a predetermined degree of separation from the first keyword node in the weighted graph based on linkage weights of links associated with user nodes and entity nodes between the first keyword node and the second keyword node.

15 . A non-transitory computer-readable medium comprising processor-executable instructions configured to cause one or more processors to:
access communication data on a communication platform associated with an enterprise; extract a plurality of keywords from the communication data; create a knowledge graph comprising a plurality of nodes, the plurality of nodes comprising a plurality of keyword nodes; identify, using a graph random walk algorithm, one or more keyword nodes in the knowledge graph related to a first keyword node; link the first keyword node and the one or more keyword nodes; determine a first keyword node embedding for the first keyword node based at least on information associated with the first keyword node and the one or more keyword nodes using a graph convolutional network algorithm; and attach the first keyword node embedding

Assignees

• Zoom Video Communications Inc

Inventors

• Han Jianbing
• Lu Ying
• Ni Kai
• Tan Jun
• Tian Wang

Classifications

• G06F16/9024

  Graphs; Linked lists (G06F16/9027 takes precedence) · CPC title

• G06F16/24549 Primary

  Run-time optimisation · CPC title

• G06F16/367 Primary

  Ontology · CPC title