Labelling Entities in a Canonical Data Model
AssigneeSap SE
InventorsNathali Ortiz Suarez, Jens Lemcke, Gunther Stuhec
Life of this patent
Estimating the term…
Abstract
Enterprises express the concepts of their electronic business-to-business (B2B) communication in diffrently structured ontology-like schemas. Collaborations benefit from merging the common concepts into semantically unique Business Entities (BEs) in a merged schema. Methods and systems for labeling the merged schema with descriptive, yet short and unique names, are described. A heuristically ranked list of descriptive candidate phrases for each BE is derived locally from the names and descriptions of the underlying concepts. A semantically unique candidate phrase is assigned to each BE that discriminates it from the other BEs by employing a constraint satisfaction problem solver.
Claims20
A method comprising:
- accessing a database that stores a merged schema based on a plurality of nodes in each schema of a plurality of schernas, the merged schema including business entity data that includes a plurality of business entities, each business entity of the plurality of business entities having a name;
- for each business entity of the plurality of business entities:
- extracting a sequence of words from the name of the business entity;
- generating candidate phrases for the business entity from the sequence of words;
- and
- ranking the candidate phrases for the business entity;
- analyzing, using a processor of a machine, candidate sets of labels for the plurality of business entities, each business entity having a label, no two business entities having the same label, the label for each business entity of the plurality of business entities being selected from the candidate phrases for the business entity;
- assigning labels to each business entity of the plurality of business entities based on the analysis of the candidate sets of labels;
- receiving data stored using a first schema of the plurality of schemas;
- converting the received data to the merged schema;
- and
- causing the presentation of the converted data using the assigned labels.
The method of claim I, wherein the ranking of the candidate phrases for the business entity includes ranking the candidate phrases for the business entity based on a frequency of words in each candidate phrase in a description of the business entity.
A system comprising:
- a schema parser module implemented by one or more processors and configured to:
- access a database that stores a merged schema based on a plurality of nodes in each schema of a plurality of schemas, the merged schema including business entity data that includes a plurality of business entities, each business entity of the plurality of business entities having a name;
- a phrase generator module implemented by the one or more processors and configured to, for each business entity of the plurality of business entities:
- extract a sequence of swords from the name of the business entity;
- generate candidate phrases for the business entity from the sequence of words;
- and
- rank the candidate phrases implemented by the one or more processors and for the business entity;
- a phrase selector module implemented by the one or more processors and configured to:
- analyze candidate sets of labels for the plurality of business entities, each business entity having a label, no two business entities having the same label, the label for each business entity of the plurality of business entities being selected from the candidate phrases for the business entity;
- and
- assign labels to each business entity of the plurality of business entities based on the analysis of the candidate sets of labels;
- and
- a conversion module implemented by the one or more processors and configured
- receive data stored using a first schema of the plurality of schemas;
- convert the received data to the merged schema;
- and
- cause a presentation of the converted data using the assigned labels.
A non-transitory machine-readable storage medium comprising instructions that, executed by one or more processors of a machine, cause the machine to perfbrm operations comprising:
- accessing a database that stores a merged schema based on a plurality of nodes in each schema of a plurality of schemas, the merged schema including business entity, data that includes a plurality of business entities, each business entity of the plurality of business entities having a name;
- for each business entity of the plurality of business entities:
- extracting a sequence of words from the name of the business entity;
- generating candidate phrases for the business entity from the sequence of words;
- and
- ranking the candidate phrases for the business entity;
- analyzing, using a processor of a machine, candidate sets of labels for the plurality of business entities, each business entity having a label, no two business entities having the same label, the label for each business entity of the plurality of business entities being selected from the candidate phrases for the business entity;
- assigning labels to each business entity of the plurality of business entities based on the analysis of the candidate sets of labels;
- receiving data stored using a first schema of the plurality of schemas;
- converting the data to the merged schema;
- and
- causing the presentation of the converted data using the assigned labels.
Dependencies are inferred from the claim text.