Browse Prior Art Database

A new approach to support the merging of multiple knowledge graphs Disclosure Number: IPCOM000031836D
Original Publication Date: 2004-Oct-13
Included in the Prior Art Database: 2004-Oct-13
Document File: 6 page(s) / 178K

Publishing Venue



Recently, KM systems have become increasingly common which provide individual knowledge graphs for all team members and extract a unified knowledge view of the whole team or organization. In order for the unified view, graphs need to be merged or aligned to one another. The processes of merging are usually handled manually and often constitute a large and tedious portion of the sharing process. We have developed a new approach based on matrix theory that supports multiple knowledge graphs merging automatically.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 43% of the total text.

Page 1 of 6

A new approach to support the merging of multiple knowledge graphs


The skill to effectively and efficiently manage knowledge is vital for sustainable corporate success. Ontology-based technologies has a high potential for user in knowledge management. According to McKinsey's estimation, business success of Ontology-based technologies will likely be a gradual process along several horizons. However, while the knowledge management software market is expected to grow from 2 billion Us $ (2001) to 6 billions in 2006, some business and technological obstacles still hinder Ontology-based technologies from becoming a true alternative to existing statistics based technologies. Basically, lots of manual works are involved in creation of the basic ontology, annotation of ontology, and merging of different ontologies.

In this invention, we provide a new approach to merge different ontologies automatically with followed characteristics:
1. To merge large numbers of ontology graphs quickly.
2. To avoid subjectivity in merging of ontologies.

Main idea:

Firstly, some preliminary definitions are introduced which are useful for subsequent discussions.

Vertex: A node in a knowledge graph which represents a knowledge point.

Edge: Direct relation between two vertexes.

Parent Vertex: start vertex in an edge.

Child Vertex: end vertex in an edge.

Access Path: If a vertex can travel thought limited edges to another vertex, we think that all the edges that it passed compose an access path.

Descendant: If there are access paths form vertex i to j, we think vertex j is a descendant of i.

Loop: In a loop, any vertex can reach all other vertexes.

Membership: avaluefrom 0 to 1 which represents a confident rate about relationship between vertexes.

Element: a matrix cell that is mapping to an edge of a graph and the value of the element is equal to the membership of the edge.

L Triangle Matrix: A lower triangular matrix with elements f[i,j] below the diagonal. The process is described as the follow graph:


Page 2 of 6

Then we will give detailed explanation for this process.

1. Define a weighted knowledge matrix for individual knowledge:

Each person owns and updates his knowledge graph. For any relationship between


[This page contains 1 picture or other non-text object]

Page 3 of 6

knowledge, the owner will assign a membership to represent the credibility of the vertexes' relationship. The knowledge graph of human k can be defined as follow:



 j i M k



 ) , (


 





The membership can be assigned by person simply or acquired through the analysis of knowledge base.

2. The combine of different knowledge matrix:

Matrix N represents a combined result from all individual knowledge matrixes. For example, we give a common method to combine different knowledge matrixes. If there are totally K persons in one team, the matrix N can be computed as follow:

   =Mi K N






3. The auxiliary matrix P based on N.

Matrix N...