Automatically generated ontology by combining structured and/or semi-structured knowledge sources

• US 7,512,576 B1
• Filed: 01/16/2008
• Issued: 03/31/2009
• Est. Priority Date: 01/16/2008
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: International Business Machines Corporation
• Inventors: Ramakrishnan, Cartic, Syeda-Mahmood, Tanveer F.
• Primary Examiner(s): Hirl; Joseph P
• Application Number:
  US12/014,876
• Time in Patent Office: 440 Days
• Field of Search: 706/45, 706/46, 706/47, 707/203
• US Class Current: 706/45
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Exemplary embodiments of the present invention relate to a solution for the extraction of information from structured and semi-structured knowledge sources. Further, ontological relationships are inferred between the extracted information by parsing the layout structure of the structured and semi-structured knowledge sources. The inferred ontological relationships are verified in relation to existing ontological sources.

Claims

1. A method for generating computer implemented ontology based upon data obtained from structured and semi-structured knowledge sources, the method comprising:

determining a primary topic concept on which an ontology is to be based;

determining, from a derivation of the Unified Medical Language System (UMLS) semantic network, ontology relationship references that are to be associated with all concepts that are comprised within the ontology;

retrieving at least one information source;

analyzing the at least one information source in order to identify a topic concept that matches the primary topic concept in addition to content that is associated with the topic concept;

identifying within the content that is associated with the topic concept, relationship references that are associated with the topic concept, wherein the relationship references respectively correspond to the determined ontology relationship references;

extracting predetermined concept descriptors from the content that is associated with the relationship references;

associating the concept descriptor with the ontology relationship reference that corresponds to the relationship reference;

analyzing the at least one information source in order to further identify concept descriptors that match the extracted concept descriptors and content that is associated with the concept descriptors;

identifying within the content that is associated with a respective concept descriptor relationship references that are associated with the concept descriptor, wherein the relationship references respectively correspond to the determined ontology relationships;

extracting concept descriptors from the content that is associated with the relationship references;

analyzing the ontology relationship references that are associated with a parent concept descriptor and a child concept descriptor;

validating the relationship between the parent concept descriptor and the child concept descriptor by capturing a corresponding relationship from the UMLS semantic network and comparing the relationship between the parent concept descriptor and the child concept descriptor to the relationship captured from the UMLS semantic network, wherein the instance that the relationship between the parent concept descriptor and the child concept descriptor is validated then the child concept descriptor is annotated as a direct ontology relationship reference of its parent concept descriptor; and

generating a primary topic concept specific ontology.

2. The method of claim 1, wherein the information source can comprise one or a combination of a structured information source and a semi-structured information source.