Unsupervised method for enriching RDF data sources from denormalized data

Patent metadata

Field	Value
Publication number	US-11068439-B2
Application number	US-201615180865-A
Country	US
Kind code	B2
Filing date	Jun 13, 2016
Priority date	Jun 13, 2016
Publication date	Jul 20, 2021
Grant date	Jul 20, 2021

Abstract

A method for unsupervised transformation of denormalized data into RDF data includes summarizing an existing resource description framework (RDF) file wherein a data summary file is produced, wherein said RDF file includes a plurality of nodes, property labels, and literals and is represented as a directed graph, and said data summary file is a graph that includes a single instance of each node-property-literal triple in the RDF file; analyzing a plurality of denormalized records wherein a record summary is produced, wherein said record summary is represented as a directed graph with a root node corresponding to each record connected to a node for each field in the record; determining one or more transformation patterns that match elements of the record summary to elements of the data summary; and transforming the plurality of denormalized records into a new RDF file by executing one of the one or more transformation patterns.

First claim

What is claimed is:

1. A computer implemented method for unsupervised transformation of denormalized data into RDF data, the method performed by the computer comprising the steps of:
summarizing an existing resource description framework (RDF) file wherein a data summary file is produced, wherein said RDF file includes a plurality of nodes that represent RDF triples, wherein RDF subjects connected to other nodes that represent RDF objects by labeled directed edges that represent RDF predicates, and said data summary file is a graph that includes, for each root node, each predicate appears once and is represented by a directed edge, and other nodes appear as necessary as sources or destinations of the edges;
analyzing a plurality of denormalized records wherein a denormalized record summary is produced, wherein denormalized records are records that have not been reduced to relational database fields and tables, wherein each denormalized record includes an identifier and a plurality of data fields associated with the identifier, wherein said denormalized record summary is represented as a plurality of directed graphs wherein a root node of each graph corresponds to the identifier of each record, and each record is connected by a directed edge to a node for each data field in the record;
determining one or more transformation patterns that match elements of the denormalized record summary to elements of the data summary; and
transforming the plurality of denormalized records into a new plurality of RDF triples by executing one of the one or more transformation patterns, wherein a subject identifier in each RDF triple of the new plurality of RDF triples is a reference to an identifier of a record of the plurality of denormalized records, and an object identifier in each RDF triple of the new plurality of RDF triples is a reference to a data field of the record of the plurality of denormalized records associated with the subject identifier of said RDF triple; and
transforming the new plurality of RDF triples into an RDF graph.

2. The method of claim 1 , further comprising creating an enriched RDF file by appending the new RDF file to the existing RDF file.

3. The method of claim 1 , further comprising ranking the one or more transformation patterns based on which transformation pattern best matches elements of the denormalized record summary to elements of the data summary.

4. The method of claim 3 , further comprising selecting one of said one of the one or more transformation patterns as a best transformation pattern, based on the rankings of the one or more transformation patterns, where said best transformation pattern comprises a plurality of records, wherein each record includes an RDF subject, an RDF predicate and an RDF object.

5. The method of claim 4 , wherein criteria for ranking the one or more transformation patterns include a size of a set of transformation patterns,
wherein criteria for selecting best transformation pattern include
which transformation pattern is a smallest set of transformation patterns that provides a best coverage of the denormalized records would be selected as the best possible pattern,
which transformation pattern provides a maximum number of columns from the denormalised data in the transformation pattern,
which transformation pattern has a maximum number of columns from the denormalised data in the transformation pattern times a number of patterns, and
which transformation pattern has a maximum number of columns from the denormalised data in the transformation times a number of patterns times a diameter of a graph of the transformation pattern.

6. The method of claim 3 , wherein each RDF triple of the new plurality of RDF triples includes a ranking.

7. The method of claim 1 , further comprising monitoring changes to the existing RDF file, and updating the data summary file when a change occurs in the existing RDF file.

8. The method of claim 1 , further comprising validating the one of the one or more transformation patterns by checking whether some selected transformation patterns are consistent with a domain knowledge of a user.

9. The method of claim 1 , wherein the direction of an arrow indicates that the value of a destination node pointed to by the arrow is associated with a label of the arrow and a value of the source node, with a relationship being denoted by the edge label, wherein leaf nodes are labeled with values, and edges are labeled with the relationship that connects a source node to the destination node.

10. The method of claim 1 , wherein transforming the new plurality of RDF triples into an RDF graph comprises following links in the new plurality of RDF triples to obtain node values and edge labels from the plurality of denormalized records
wherein each root nodes is labeled with the identifier of one of the denormalized records of the plurality of denormalized records, values of leaf nodes associated with each root node correspond to data fields associated with the identifier associated with the root node, and each edge is labeled with a predicate value of the RDF triple whose subject corresponds to a source node of said edge.

11. A non-transitory program storage device readable by a computer, tangibly embodying a program of instructions executed by the computer to perform the method steps for unsupervised transformation of denormalized data into RDF data, the method comprising the steps of:
summarizing an existing resource description framework (RDF) file wherein a data summary file is produced, wherein said RDF file includes a plurality of nodes that represent RDF triples, wherein RDF subjects connected to other nodes that represent RDF objects by labeled directed edges that represent RDF predicates, and said data summary file is a graph that includes, for each root node, each predicate appears once and is represented by a directed edge, and other nodes appear as necessary as sources or destinations of the edges;
analyzing a plurality of denormalized records wherein a denormalized record summary is produced, wherein denormalized records are records that have not been reduced to relational database fields and tables, wherein each denormalized record includes an identifier and a plurality of data fields associated with the identifier, wherein said denormalized record summary is represented as a plurality of directed graphs wherein a root node of each graph corresponds to the identifier of each record, and each record is connected by a directed edge to a node for each data field in the record;
determining one or more transformation patterns that match elements of the denormalized record summary to elements of the data summary; and
transforming the plurality of denormalized records into a new plurality of RDF triples by executing one of the one or more transformation patterns, wherein a subject identifier in each RDF triple of the new plurality of RDF triples is a reference to an identifier of a record of the plurality of denormalized records, and an object identifier in each RDF triple of the new plurality of RDF triples is a reference to a data field of the record of the plurality of denormalized records associated with the subject identifier of said RDF triple; and
transforming the new plurality of RDF triples into an RDF graph.

12. The non-transitory program storage device of claim 11 , the method further comprising creating an enriched RDF file by appending the new RDF file to the existing RDF file.

13. The non-transitory program storage device of claim 11 , the method further comprising ranking the one or more transformation patterns based on which transformation pattern best matches elements of

Assignees

• IBM

Inventors

• Bicer Veli
• LOPES Nuno
• Kotoulas Spyros
• Stephenson Martin

Classifications

• G06F16/9024

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

• G06F16/116 Primary

  Details of conversion of file system types or formats · CPC title

• G06F16/2365

  Ensuring data consistency and integrity · CPC title

• G06F16/84

  Mapping; Conversion · CPC title