US 8,620,964 B2Granted

Ontology construction

AssigneeMotorola Mobility LLC

InventorsDorothea Tsatsou, Paul C. Davis, Jianguo Li, Ioannis Kompatsiaris, Symeon Papadopoulos

Life of this patent

Estimating the term…

Loading patent…

Abstract

Disclosed are methods in which ontologies are automatically (i.e., with little or no human interaction) constructed from data. The constructed ontology may be provided for use by an application or device. An ontology may be constructed as follows. Firstly, a lexical graph is constructed. Secondly, a concept-detection process is performed on the lexical graph to detect concepts (prominent terms) in the lexical graph. Thirdly, a clustering method is performed on the lexical graph to form groups of concepts. Fourthly, relationships between the concepts are determined based on the interconnections of the concepts in the groups. Lastly, the concepts and detected relationships therebetween are structured using a formal ontology specification.

Claims19

Dependent claims

A method of constructing an ontology from data, the ontology comprising concepts and relationships between certain of those concepts, and the ontology having a defined structure, the method comprising:

using the data, constructing, by one or more processors, a lexical graph, wherein:
the lexical graph comprises a plurality of nodes and a plurality of edges;
each node of the lexical graph corresponds to one or more data terms;
each edge of the lexical graph is a connection between two different nodes;
and
the lexical graph further comprises statistical information relating to the data terms;
selecting, by the one or more processors, as the concepts of the ontology, nodes in the lexical graph that correspond to prominent data terms;
wherein the step of selecting, as the concepts of the ontology, nodes in the lexical graph corresponding to prominent data terms comprises:
for each node in the lexical graph, determining a weight of significance;
and identifying as a node that corresponds to a prominent data term, those nodes that have a weight of significance greater than or equal to a first predetermined threshold value;
grouping, by the one or more processors, into one or more groups, the concepts, each group comprising one or more concepts and the grouping comprises performing a community-detection process based on structural similarity of nodes in the lexical graph;
based on the one or more groups and the statistical information, detecting, by the one or more processors, relationships between the concepts;
and
using a formal ontological specification, the concepts, and the relationships between those concepts, defining, by the one or more processors, the structure of the ontology.

11 more elements

Dependencies are inferred from the claim text.