Cognitive operations based on empirically constructed knowledge graphs

Patent metadata

Field	Value
Publication number	US-10664757-B2
Application number	US-201514855461-A
Country	US
Kind code	B2
Filing date	Sep 16, 2015
Priority date	Sep 16, 2015
Publication date	May 26, 2020
Grant date	May 26, 2020

Abstract

Mechanisms are provided for performing a cognitive operation. The mechanisms receive an original graph data structure comprising nodes and edges between nodes and activity log information for nodes of the original graph data structure. The mechanisms identify a set of nodes in the original graph data structure having a predetermined pattern of activity in the activity log information, and a set of edges between these nodes. The mechanisms calculate an importance weight for each edge in the set of edges and modify the original graph data structure based on the calculated importance weights for the edges in the set of edges, to thereby generate a modified graph data structure. The mechanisms then perform a cognitive operation based on the modified graph data structure. The set of edges may comprise actual edges between the nodes and/or potential edges between the nodes.

First claim

What is claimed is:

1. A method, in a data processing system comprising a processor and a memory, for performing a cognitive operation, the method comprising:
receiving, by the data processing system, an original graph data structure comprising nodes and edges between nodes;
receiving, by the data processing system, activity log information for nodes of the original graph data structure, wherein an activity metric associated with a node, in the activity log information, represents activities performed with regard to the object itself that is represented by the node;
identifying, by the data processing system, a set of nodes in the original graph data structure having a predetermined pattern of activity in the activity log information, and identifying a set of edges between these nodes;
evaluating the predetermined patterns of activity associated with the nodes in the set of nodes to identify correlations between activity of first nodes in the set of nodes with activity of other second nodes in the set of nodes;
calculating, by the data processing system, an importance weight for each edge in the set of edges based on results of evaluating the predetermined patterns of activity to identify correlations between activity of the first nodes in the set of nodes with other second nodes in the set of nodes;
modifying, by the data processing system, the original graph data structure based on the calculated importance weights for the edges in the set of edges at least by comparing the calculated importance weights for the edges to at least one noise criterion, to thereby generate a modified graph data structure; and
performing, by the data processing system, a cognitive operation based on the modified graph data structure, wherein the set of edges comprises at least one of actual edges between the nodes and potential edges between the nodes, wherein modifying the original graph data structure to generate the modified graph data structure comprises removing one or more of the edges in the set of edges from the original graph data structure, where the removed edge is determined to be noise in the original graph data structure based on the at least one noise criterion.

2. The method of claim 1 , wherein the subset of nodes of the graph are nodes determined to be associated with one or more popular nodes identified by the predetermined pattern of activity, wherein a node is a popular node when an activity metric of the node exceeds a threshold level of activity repeatedly within a predetermined period of time.

3. The method of claim 2 , wherein the identified set of edges are popular-to-popular edges that connect a first popular node to a second popular node.

4. The method of claim 3 , further comprising calculating, for each popular-to-popular edge in the set of edges, a correlation metric that correlates an activity pattern in the activity log information for the first popular node with an activity pattern in the activity log information for the second popular node of the popular-to-popular edge, and wherein modifying the original graph data structure based on the calculated importance weights for the edges in the set of edges comprises modifying the original graph data structure based on the calculated correlation metrics of each of the popular-to-popular edges in the set of edges.

5. The method of claim 4 , wherein modifying the original graph data structure based on the calculated correlation metrics of each of the popular-to-popular edges in the set of edges comprises, for each of the popular-to-popular edges;
determining if a correlation metric for the popular-to-popular edge satisfies a predetermined relationship with a predetermined correlation threshold value; and
in response to determining that the correlation metric for the popular-to-popular edge does not satisfy the predetermined relationship, removing the popular-to-popular edge from the original graph data structure when generating the modified graph data structure based on the original graph data structure.

6. The method of claim 1 , wherein the at least one noise criterion comprises a predetermined correlation threshold value, and wherein modifying the original graph data structure based on the calculated importance weights for the edges in the set of edges comprises:
calculating, for each edge in the set of edges, a correlation metric that correlates activity information for the nodes connected by the edge, based on the importance weight associated with the edge;
determining, for each edge in the set of edges, whether a corresponding correlation metric of the edge satisfies a predetermined relationship with a predetermined correlation threshold value; and
in response to determining that the correlation metric for the edge does not satisfy the predetermined relationship, removing the edge from the original graph data structure when modifying the original graph data structure to generate the modified graph data structure.

7. The method of claim 1 , wherein the cognitive operation is at least one of a natural language question answering operation utilizing the modified graph data structure to identify related concepts in a corpus of information, or an information retrieval operation that retrieves information and ranks the retrieved information based on the modified graph data structure.

8. The method of claim 1 , wherein the data processing system implements a parallel architecture having a plurality of processors, and wherein calculating the importance weight for each edge in the set of edges comprises distributing the calculations across the plurality of processors, and wherein modifying the original graph data structure comprises inputting, to a trimming processor in the data processing system, the importance weights calculated by the plurality of processors and modifying the original graph data structure to generate the modified graph data structure by removing edges in the set of edges from the original graph data structure that have importance weights that are below a predetermined threshold value.

9. The method of claim 1 , wherein the nodes represent web pages of one or more web sites, and wherein the activity log information stores information regarding a number of page views of one or more web pages represented by the nodes.

10. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a data processing system, causes the data processing system to:
receive an original graph data structure comprising nodes and edges between nodes;
receive activity log information for nodes of the original graph data structure, wherein an activity metric associated with a node, in the activity log information, represents activities performed with regard to the object itself that is represented by the node;
identify, by the data processing system, a set of nodes in the original graph data structure having a predetermined pattern of activity in the activity log information, and identify a set of edges between these nodes;
evaluate the predetermined patterns of activity associated with the nodes in the set of nodes to identify correlations between activity of first nodes in the set of nodes with activity of other second nodes in the set of nodes;
calculate an importance weight for each edge in the set of edges based on results of evaluating the predetermined patterns of activity to identify correlations between activity of the first nodes in the set of nodes with other second nodes in the set of nodes;
modify the original graph data structure based on the calculated importance weights for the edges in the set of edges at least by comparing the calcula

Assignees

• IBM

Inventors

• Lastras-Montano Luis A
• Misra Vinith
• Soares Livio B

Classifications

• G06F16/3344

  using natural language analysis · CPC title

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06F16/3329

  Natural language query formulation · CPC title

• G06F16/9024

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