Automatic generation of data objects from user input

Patent metadata

Field	Value
Publication number	US-12423311-B1
Application number	US-202418893785-A
Country	US
Kind code	B1
Filing date	Sep 23, 2024
Priority date	Sep 23, 2024
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

The present disclosure relates to techniques for automatically generating new data objects from user input. The system receives user input comprising a plurality of words and executes a first query on a vector store to identify schema elements similar to keywords in the user input. The vector store provides a response with similarity scores for identified elements. A second query is executed on a knowledge graph to identify association paths between data objects that include the identified elements. The knowledge graph response includes association information linking source and target data objects through selected elements. Full association paths are constructed from this information, and a command is generated to instantiate a new data object with elements corresponding to the user input. This approach leverages the strengths of large language models, vector stores, and knowledge graphs to efficiently and accurately create new data objects, ensuring data integrity and relevance.

First claim

What is claimed is:

1. A computing system comprising:
at least one hardware processor;
at least one memory coupled to the at least one hardware processor; and
one or more computer-readable storage media comprising computer-executable instructions that, when executed, cause the computing system to perform operations comprising:
receiving user input comprising a plurality of words;
causing a first query to be executed on a vector store to identify elements of a schema, the schema comprising definitions of a plurality of data objects that each include a plurality of elements that are similar to keywords in the user input, where the vector store provides a first query response that comprises similarity scores for identified elements of the schema;
causing a second query to be executed on a knowledge graph to identify association paths between at least a portion of data objects of the schema having definitions that comprise at least one element of the one or more of the identified elements, where a second query response comprises association information for given pairs of two or more pairs of data objects of the at least a portion of the data objects of the schema, the association information comprising a source data object, a target data object, and at least one of the one or more selected elements that serve to operationally link the source data object and the target data object;
from the association information, constructing full association paths between data objects of the at least a portion of the data object; and
from the full association paths, generating a command to instantiate a data object having elements corresponding to keywords of the user prompt.

2. The computing system of claim 1 , the operations further comprising:
submitting a first prompt to a large language model, the first prompt comprising at least a portion of the identified elements with an instruction to select identified elements that are relevant to the user input, wherein a first prompt response from the large language model comprises one or more selected elements selected from the at least a portion of the identified elements, where the second query is generated using the selected elements.

3. The computing system of claim 2 , the operations further comprising:
sorting execution results of the first query by similarity score to provide sorted elements, wherein the first prompt comprises the at least a portion of the identified elements as sorted elements.

4. The computing system of claim 1 , the operations further comprising:
processing the knowledge graph to provide extracted information regarding data objects and data object elements in the knowledge graph; and
generating semantic embeddings for the extracted information; and
storing the semantic embeddings in the vector store.

5. The computing system of claim 1 , the operations further comprising:
receiving an identifier for a starting data object of the schema; and
causing a third query to be executed on the knowledge graph to identify data objects being directly or indirectly related to the starting data object, where data objects used in the second query are data objects identified by the third query.

6. The computing system of claim 5 , wherein the identifier is provided with the user input.

7. The computing system of claim 5 , the operations further comprising:
causing a third query to be executed on the vector store to identify elements of the schema that are similar to keywords of the user prompt, where the vector store provides a third query response that comprises similarity scores for identified elements of the schema, where the third query is the first query or is a different query;
causing a fourth query to be executed on the knowledge graph to identify data objects of the schema to determine if an element of the identified elements of the schema of the third query response is included in a given data object of the schema or can be retrieved via associations of the given data object with other data objects of the schema, the fourth query providing a fourth query response comprising data objects identified by the fourth query;
for data objects identified in the fourth query response, determining a number of elements reachable through a given data object of the data objects identified in the fourth query response; and
selecting as the starting data object a data object having a highest number of elements.

8. The computing system of claim 5 , wherein the first query is constrained to a starting data object and data objects that are directly or indirectly related to the starting data object.

9. The computing system of claim 5 , wherein first queries are performed for the starting data object and each data object that is directly or indirectly related to the starting data object.

10. The computing system of claim 5 , the operations further comprising:
causing a third query to be executed on the vector store to identify elements of the schema that are similar to keywords of the user prompt, where the vector store provides a third query response that comprises similarity score for identified elements of the schema, where the third query is the first query or is different than the first query;
causing a fourth query to be executed on the knowledge graph to identify data objects of the schema to determine if an element of the identified elements of the schema of the third query response is included in a given data object of the schema or can be retrieved via associations of the given data object with other data objects of the schema, the fourth query providing a fourth query response comprising data objects identified by the fourth query;
for data objects identified in the fourth query response, determining a number of elements reachable through a given data object of the data objects identified in the fourth query response;
determining that multiple data objects identified in the fourth query response have a highest number of reachable elements;
displaying to a user identifiers for data objects of the multiple data objects; and
receiving user input selecting a data object of the multiple data objects as the starting data object.

11. The computing system of claim 1 , the operations further comprising:
submitting a first prompt to a large language model, the first prompt comprising the user input and an instruction to identify the keywords in the user input.

12. The computing system of claim 1 , the operations further comprising:
submitting a first prompt to a large language model, the first prompt comprising the full association paths and an instruction to select a full association path the best matches the user input.

13. The computing system of claim 1 , the operations further comprising:
causing a third query to be executed on the vector store to evaluate similarity of the full association paths to the user input.

14. The computing system of claim 13 , the operations further comprising:
sorting execution results of the third query by similarity score to provide sorted full association paths; and
submitting a first prompt to a large language model, the first prompt comprising the sorted full association paths and an instruction to select a full association path that best matches the user input.

15. The computing system of claim 1 , the operations further comprising:
processing the knowledge graph to provide extracted information regarding data objects and data object elements in the knowledge graph;
generating semantic embeddings for the extracted information;
storing the semantic embeddings in the vector store;
submitting a first promp

Assignees

• Sap Se

Inventors

• Stoll Sascha
• Droll Sebastian
• Balzar Andreas
• Buchmann Daniel
• Krannich Bernd

Classifications

• G06F16/212 Primary

  with details for data modelling support · CPC title

• G06F16/24578 Primary

  using ranking · CPC title

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title