Granted Patent B2

US 12,400,145 · App. 17/715,487 · Granted Aug 26, 2025

Inventors: Zhiqiang Yuan (San Jose, CA); Yujing Qian (Mountain View, CA)

Assignee: Deere & Company

G06N20/00

Abstract

Claims (63)

1. A method implemented by one or more processors, the method comprising:

identifying a table of agricultural data, where the table of agricultural data includes a plurality of agricultural records, and where each agricultural record includes one or more agricultural attributes;

processing the table of agricultural data using a mapping machine learning model to generate a relationship embedding indicating a relationship between each of the one or more agricultural attributes and one or more nodes in a knowledge graph, where each node indicates a mapping agricultural attribute, where each node is connected with one or more additional nodes in the knowledge graph, and where the connection between each node and the one or more additional nodes represents a relationship between the node and the one or more additional nodes; and

causing a computing device to perform one or more actions based on the relationship embedding, wherein causing the computing device to perform the one or more actions based on the relationship embedding includes:

generating a normalized table of agricultural data based on the table of agricultural data and the relationship embedding; and

processing the normalized table of agricultural data using an additional machine learning model to generate one or more agricultural predictions.

2. The method of claim 1 , further including:

identifying an additional table of agricultural data, wherein the additional table of agricultural data includes a plurality of additional agricultural records, wherein each additional agricultural record corresponds to one or more additional agricultural attributes, and wherein at least one of the one or more additional agricultural attributes is distinct from the one or more agricultural attributes;

processing the additional table of agricultural data using the mapping machine learning model to generate an additional relationship embedding indicating an additional relationship between each of the one or more additional agricultural attributes and a plurality of nodes of the knowledge graph; and

causing the computing device to perform one or more additional actions based on the additional relationship embedding.

3. The method of claim 2 , wherein one of the one or more agricultural attributes includes a dry yield, wherein one of the one or more additional agricultural attributes includes a harvested yield, wherein a dry yield node is connected with a harvested yield node via one or more edges of the knowledge graph.

4. The method of claim 1 , wherein processing the table of agricultural data using the mapping machine learning model to generate the relationship embedding includes:

processing the table of agricultural data using an embedding model portion of the mapping machine learning model to generate an embedding space representation of the one or more agricultural attributes;

identifying one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on a distance between the embedding space representation of a corresponding agricultural attribute and an embedding space representation of each of the one or more nodes; and

generating the relationship embedding based on the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes.

5. The method of claim 4 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes satisfies a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the distance between the embedding space representation of each of the one or more nodes satisfies the threshold value, identifying the one or more nodes of the knowledge graph that correspond to the one or more agricultural attributes based on a node with the shortest distance to the embedding space representation of the corresponding agricultural attribute.

6. The method of claim 4 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy the threshold value, generating an additional node based on the embedding space representation of the corresponding agricultural attribute, where the additional node corresponds to the corresponding agricultural attribute.

7. A computer program comprising instructions that when executed by one or more processors of a computing system, cause the computing system to perform a method of:

identifying a table of agricultural data, where the table of agricultural data includes a plurality of agricultural records, and where each agricultural record includes one or more agricultural attributes;

processing the table of agricultural data using a mapping machine learning model to generate a relationship embedding indicating a relationship between each of the one or more agricultural attributes and one or more nodes in a knowledge graph, where each node indicates a mapping agricultural attribute, where each node is connected with one or more additional nodes in the knowledge graph, and where the connection between each node and the one or more additional nodes represents a relationship between the node and the one or more additional nodes; and

causing a computing device to perform one or more actions based on the relationship embedding, wherein causing the computing device to perform the one or more actions based on the relationship embedding includes:

generating a normalized table of agricultural data based on the table of agricultural data and the relationship embedding; and

processing the normalized table of agricultural data using an additional machine learning model to generate one or more agricultural predictions.

8. The computer program of claim 7 , wherein the instructions further include:

identifying an additional table of agricultural data, wherein the additional table of agricultural data includes a plurality of additional agricultural records, wherein each additional agricultural record corresponds to one or more additional agricultural attributes, and wherein at least one of the one or more additional agricultural attributes is distinct from the one or more agricultural attributes;

processing the additional table of agricultural data using the mapping machine learning model to generate an additional relationship embedding indicating an additional relationship between each of the one or more additional agricultural attributes and a plurality of nodes of the knowledge graph; and

causing the computing device to perform one or more additional actions based on the additional relationship embedding.

9. The computer program of claim 8 , wherein one of the one or more agricultural attributes includes a dry yield, wherein one of the one or more additional agricultural attributes includes a harvested yield, wherein a dry yield node is connected with a harvested yield node via one or more edges of the knowledge graph.

10. The computer program of claim 7 , wherein processing the table of agricultural data using the mapping machine learning model to generate the relationship embedding includes:

processing the table of agricultural data using an embedding model portion of the mapping machine learning model to generate an embedding space representation of the one or more agricultural attributes;

identifying one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on a distance between the embedding space representation of a corresponding agricultural attribute and an embedding space representation of each of the one or more nodes; and

generating the relationship embedding based on the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes.

11. The method of claim 10 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes satisfies a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the distance between the embedding space representation of each of the one or more nodes satisfies the threshold value, identifying the one or more nodes of the knowledge graph that correspond to the one or more agricultural attributes based on a node with the shortest distance to the embedding space representation of the corresponding agricultural attribute.

12. The method of claim 10 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy the threshold value, generating an additional node based on the embedding space representation of the corresponding agricultural attribute, where the additional node corresponds to the corresponding agricultural attribute.

13. A non-transitory computer-readable storage medium storing instructions executable by one or more processors of a computing system to perform a method of:

identifying a table of agricultural data, where the table of agricultural data includes a plurality of agricultural records, and where each agricultural record includes one or more agricultural attributes;

processing the table of agricultural data using a mapping machine learning model to generate a relationship embedding indicating a relationship between each of the one or more agricultural attributes and one or more nodes in a knowledge graph, where each node indicates a mapping agricultural attribute, where each node is connected with one or more additional nodes in the knowledge graph, and where the connection between each node and the one or more additional nodes represents a relationship between the node and the one or more additional nodes; and

causing a computing device to perform one or more actions based on the relationship embedding, wherein causing the computing device to perform the one or more actions based on the relationship embedding includes:

generating a normalized table of agricultural data based on the table of agricultural data and the relationship embedding; and

processing the normalized table of agricultural data using an additional machine learning model to generate one or more agricultural predictions.

14. The non-transitory computer readable medium of claim 13 , wherein the instructions further include:

identifying an additional table of agricultural data, wherein the additional table of agricultural data includes a plurality of additional agricultural records, wherein each additional agricultural record corresponds to one or more additional agricultural attributes, and wherein at least one of the one or more additional agricultural attributes is distinct from the one or more agricultural attributes;

processing the additional table of agricultural data using the mapping machine learning model to generate an additional relationship embedding indicating an additional relationship between each of the one or more additional agricultural attributes and a plurality of nodes of the knowledge graph; and

causing the computing device to perform one or more additional actions based on the additional relationship embedding.

15. The non-transitory computer readable medium of claim 14 , wherein one of the one or more agricultural attributes includes a dry yield, wherein one of the one or more additional agricultural attributes includes a harvested yield, wherein a dry yield node is connected with a harvested yield node via one or more edges of the knowledge graph.

16. The non-transitory computer readable medium of claim 13 , wherein processing the table of agricultural data using the mapping machine learning model to generate the relationship embedding includes:

processing the table of agricultural data using an embedding model portion of the mapping machine learning model to generate an embedding space representation of the one or more agricultural attributes;

identifying one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on a distance between the embedding space representation of a corresponding agricultural attribute and an embedding space representation of each of the one or more nodes; and

generating the relationship embedding based on the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes.

17. The non-transitory computer readable medium of claim 16 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes satisfies a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the distance between the embedding space representation of each of the one or more nodes satisfies the threshold value, identifying the one or more nodes of the knowledge graph that correspond to the agricultural attribute based on a node with the shortest distance to the embedding space representation of the corresponding agricultural attribute.

18. The non-transitory computer readable medium of claim 16 , wherein identifying the one or more nodes of the knowledge graph corresponding to each of the one or more agricultural attributes based on the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes includes:

determining whether the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy a threshold value; and

in response to determining the distance between the embedding space representation of the corresponding agricultural attribute and the embedding space representation of each of the one or more nodes fails to satisfy the threshold value, generating an additional node based on the embedding space representation of the corresponding agricultural attribute, where the additional node corresponds to the corresponding agricultural attribute.

Assignments (3)

Continuity (1)

References Cited (14)