Adaptive evaluation of meta-relationships in semantic graphs

Patent metadata

Field	Value
Publication number	US-11270076-B2
Application number	US-201715844723-A
Country	US
Kind code	B2
Filing date	Dec 18, 2017
Priority date	Jun 26, 2017
Publication date	Mar 8, 2022
Grant date	Mar 8, 2022

Abstract

A method and system are provided for adaptive evaluation of meta-relationships in semantic graphs. The method includes providing a semantic graph based on a knowledge base in which concepts in the form of graph nodes are linked by semantic relationships in the form of graph edges. Metadata are encoded in the edges and nodes of the semantic graph, of weightings for measuring a meta-relationship, wherein the meta-relationship applies to the concepts of the semantic graph and is independent of the semantic relationship defined by the edges of the semantic graph. A graph activation is carried out for an input context relating to one or more concepts of the semantic graph, wherein the weightings are applied to a spreading activation signal through the semantic graph to produce a measure of the meta-relationship for a sub-set of concepts of the semantic graph.

First claim

What is claimed is:

1. A computer-implemented method for adaptive evaluation of meta-relationships in semantic graphs, the method comprising:
deriving a semantic graph from a natural language source based on a knowledge base in which concepts in a form of graph nodes are linked by semantic relationships in a form of graph edges;
encoding in metadata of the edges and nodes of a semantic graph, weightings as feature vectors for measuring a meta-relationship, wherein the meta-relationship applies to the concepts of the semantic graph and is independent of the semantic relationship defined by the edges of the semantic graph, wherein the feature vectors comprise an intensity of the meta-relationship and a confidence score of the intensity, and wherein the intensity uses runtime inputs for the nodes;
carrying out a graph activation for an input context relating to one or more concepts of the semantic graph, wherein the weightings as the feature vectors are applied to a spreading activation signal through the semantic graph to produce a measure of the meta-relationship for a sub-set of concepts of the semantic graph;
scoring the edges and nodes of the semantic graph in response to the graph activation;
seeding the graph with seed weightings for measures of the meta-relationship obtained from a set of resources independent of the knowledge base on which the semantic graph is based; and
in response to additions to the set of resources independent of the knowledge base on which the semantic graph is based, updating the seed weightings.

2. The method of claim 1 , wherein carrying out a graph activation for the input context further comprises:
discovering instances of concepts in the input context;
activating nodes corresponding to the concepts in the semantic graph;
traversing a signal outward to adjacent nodes;
activating the adjacent nodes in turn whilst applying the weightings as the feature vectors to the signal; and
determining one or more focus nodes with a highest resultant activation signal.

3. The method of claim 2 , further comprising outputting a resultant activated portion of the semantic graph reflecting a measurement of the meta-relationship in the input context.

4. The method of claim 1 , wherein the weightings as the feature vectors for measures of the meta-relationship are raw values that are obtained from the set of resources and are updated in response to additions to the set of resources, wherein the raw values are applied during the graph activation.

5. The method of claim 1 , wherein the weightings as the feature vectors indicate multi-dimensional measurements for different aspects of the meta-relationship and polarities of the meta-relationship.

6. The method of claim 1 , wherein the weightings as the feature vectors are calculated in response to runtime inputs for the nodes for instances of concepts of the input context.

7. The method of claim 6 , wherein the feature vectors include relevance factors to be applied to the runtime inputs for the nodes.

8. The method of claim 7 , wherein the relevance factors are different for different nodes.

9. The method of claim 6 , wherein the feature vectors include semantic and lexical features for instances of concepts in the input context in addition to the meta-relationship.

10. The method of claim 6 , wherein the feature vectors define confidence scores for the weightings as the feature vectors.

11. The method of claim 6 , wherein the feature vectors define aggregation of meta-relationship measures using statistical techniques.

12. The method of claim 1 , wherein the meta-relationship relates to a phenomenon in a form of one of a group consisting of:
sentiment analysis, bias evaluation, bias in predictive analysis, query expansion using information retrieval, risk assessment, geo-spatial inference, and suitability of treatment, use or handling including clinical trial matching.

Assignees

• IBM

Inventors

• Mcateer Seamus R
• Mccloskey Daniel
• Mohan Aditya
• Sogrin Mikhail

Classifications

• G16H10/20

  for electronic clinical trials or questionnaires · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06F16/9024

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

• G06F16/3338 Primary

  Query expansion · CPC title