Back to results
US 11,386,335 B2Granted
Systems and methods providing evolutionary generation of embeddings for predicting links in knowledge graphs
AssigneeAccenture Global Solutions Limited
InventorsChristophe Gueret, Luca Costabello
Life of this patent
Estimating the term…
Abstract
Complex computer system architectures are described for analyzing data elements of a knowledge graph, and predicting new facts from relational learning applied to the knowledge graph. This discovery process includes converting the knowledge graph into a set of candidate embeddings spaces to apply further analysis to rank the set of candidate embeddings spaces, where the top ranked candidate embeddings spaces are further processed to identify the new facts.
Claims20
A computing device comprising:
- a data reception circuitry configured to receive a knowledge graph including a set of structured data;
- a knowledge graph embedding circuitry configured to:
- select a sampled set of vector triples from the knowledge graph;
- convert the knowledge graph to a set of candidate embeddings spaces, wherein each of the candidate embeddings spaces includes a different set of vectors representing the sampled set of vector triples from the knowledge graph;
- and
- a processing circuitry configured to:
- calculate a weighted mean reciprocal rank of the sampled set of vector triples for each of the candidate embeddings spaces;
- determine a respective fact score for each of the candidate embeddings spaces based on the calculated weighted mean reciprocal rank;
- rank the candidate embeddings spaces according to their respective fact scores;
- and
- select a predetermined number of the highest ranked candidate embeddings spaces when a stopping condition is met.
8 more elements
A method comprising:
- receiving, by a reception circuitry, a knowledge graph including a set of structured data;
- selecting, by a knowledge graph embedding circuitry, a sampled set of vector triples from the knowledge graph;
- converting, by the knowledge graph embedding circuitry, the knowledge graph to a set of candidate embeddings spaces, wherein each of the candidate embeddings spaces includes a different set of vectors representing the sampled set of vector triples from the knowledge graph;
- calculating, by a processing circuitry, a weighted mean reciprocal rank of the sampled set of vector triples for each of the candidate embeddings spaces;
- determining, by the processing circuitry, a respective fact score for each of the candidate embeddings spaces based on the calculated weighted mean reciprocal rank;
- ranking, by the processing circuitry, the candidate embeddings spaces according to their respective fact scores;
- and
- selecting, by the processing circuitry, a predetermined number of the highest ranked candidate embeddings spaces when a stopping condition is met.
5 more elements
A computing device comprising:
- a processing circuitry;
- and
- a non-transitory storage medium configured to store instructions that, when executed, causes the processing circuitry to:
- receive a knowledge graph including a set of structured data;
- select a sampled set of vector triples from the knowledge graph;
- convert the knowledge graph to a set of candidate embeddings spaces, wherein each of the candidate embeddings spaces includes a different set of vectors representing the sampled set of vector triples from the knowledge graph;
- calculate a weighted mean reciprocal rank of the sampled set of vector triples for each of the candidate embeddings spaces;
- determine a respective fact score for each of the candidate embeddings spaces based on the calculated weighted mean reciprocal rank;
- rank the candidate embeddings spaces according to their respective fact scores;
- and
- select a predetermined number of the highest ranked candidate embeddings spaces when a stopping condition is met.
8 more elements
Dependencies are inferred from the claim text.