Back to results
US 2024/0256771 A1Published application
Training graph neural network to identify key-value pairs in documents
AssigneeOracle International Corporation
InventorsXu Zhong, Thanh Long Duong, Mark Johnson
Life of this patent
Estimating the term…
Abstract
Techniques for identifying content in key-value pairs of documents using a graph neural network (GNN) are disclosed. A system trains a GNN to identify key-value pair groupings in documents. The GNN classifies nodes in document graphs as key-type nodes and answer-type nodes. The GNN also classifies edges connecting nodes in the document graphs for keeping in the document graph or removing from the document graph. The resulting document graph includes key-value groupings in the document. Upon identifying content matching a query, a system returns as a query response content from among the key-value groupings.
Claims21
One or more non-transitory computer readable media comprising instructions which, when executed by one or more hardware processors, causes performance of operations comprising:
- training a machine learning model to generate predictions identifying key-value pair groupings in documents, the training comprising:
- accessing a training data set of document graphs describing a corresponding set of documents, each respective document graph in the training data set comprising:
- a first node corresponding to a first grammatical element in a respective document of the set of documents, the first grammatical element corresponding to a key of a key-value pair grouping, the first node comprising a first embedding comprising a first set of vectors encoding a first set of attributes representing the first grammatical element in the first document;
- a second node corresponding to a second grammatical element in the document, the second grammatical element corresponding to a value of the key-value pair grouping, the second node comprising a second embedding comprising a second set of vectors encoding a second set of attributes representing the second grammatical element in the first document;
- and
- a first edge connecting the first node and the second node, the first edge comprising a third embedding comprising a third set of vectors encoding a third set of attributes representing a relationship of the first node to the second node;
- training the machine learning model based on the training data set to generate:
- based at least on values of the first set of vectors, the second set of vectors, and the third set of vectors:
- a classification of the first node as a key-type node, the second node as a value-type node, and the first edge as a positive-type edge connecting two nodes of a same key-value pair grouping;
- and
- applying the trained machine learning model to a target document graph, representing a target document, to identify one or more key-value pair groupings in the target document.
8 more elements
A method comprising:
- training a machine learning model to generate predictions identifying key-value pair groupings in documents, the training comprising:
- accessing a training data set of document graphs describing a corresponding set of documents, each respective document graph in the training data set comprising:
- a first node corresponding to a first grammatical element in a respective document of the set of documents, the first grammatical element corresponding to a key of a key-value pair grouping, the first node comprising a first embedding comprising a first set of vectors encoding a first set of attributes representing the first grammatical element in the first document;
- a second node corresponding to a second grammatical element in the first document, the second grammatical element corresponding to a value of the key-value pair grouping, the second node comprising a second embedding comprising a second set of vectors encoding a second set of attributes representing the second grammatical element in the first document;
- and
- a first edge connecting the first node and the second node, the first edge comprising a third embedding comprising a third set of vectors encoding a third set of attributes representing a relationship of the first node to the second node;
- training the machine learning model based on the training data set to generate:
- based at least on values of the first set of vectors, the second set of vectors, and the third set of vectors:
- a classification of the first node as a key-type node, the second node as a value-type node, and the first edge as a positive-type edge connecting two nodes of a same key-value pair grouping;
- and
- applying the trained machine learning model to a target document graph, representing a target document, to identify one or more key-value pair groupings in the target document.
8 more elements
A system comprising:
- one or more processors;
- and
- memory storing instructions that, when executed by the one or more processors, cause the system to perform:
- training a machine learning model to generate predictions identifying key-value pair groupings in documents, the training comprising:
- accessing a training data set of document graphs describing a corresponding set of documents, each respective document graph in the training data set comprising:
- a first node corresponding to a first grammatical element in a respective document of the set of documents, the first grammatical element corresponding to a key of a key-value pair grouping, the first node comprising a first embedding comprising a first set of vectors encoding a first set of attributes representing the first grammatical element in the first document;
- a second node corresponding to a second grammatical element in the first document, the second grammatical element corresponding to a value of the key-value pair grouping, the second node comprising a second embedding comprising a second set of vectors encoding a second set of attributes representing the second grammatical element in the first document;
- and
- a first edge connecting the first node and the second node, the first edge comprising a third embedding comprising a third set of vectors encoding a third set of attributes representing a relationship of the first node to the second node;
- training the machine learning model based on the training data set to generate:
- based at least on values of the first set of vectors, the second set of vectors, and the third set of vectors:
- a classification of the first node as a key-type node, the second node as a value-type node, and the first edge as a positive-type edge connecting two nodes of a same key-value pair grouping;
- and
- applying the trained machine learning model to a target document graph, representing a target document, to identify one or more key-value pair groupings in the target document.
11 more elements
Dependencies are inferred from the claim text.