Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Micro rule learning model for mining rules with different lengths in knowledge graph

A technology of knowledge graph and learning model, applied in the field of differentiable rule learning model, which can solve the problems of DRUM model, such as allocation reliability of difficult new relations, wrong rules, difficult learning rules, etc.

Pending Publication Date: 2022-05-27
UNIV OF SCI & TECH OF CHINA
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, experiments demonstrate that the DRUM model has difficulty assigning appropriate confidences to new relations, making it difficult to learn rules of different lengths
The Neural LP model uses an attention mechanism to learn rules of different lengths, however, it often introduces false rules when mining longer rules

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Micro rule learning model for mining rules with different lengths in knowledge graph
  • Micro rule learning model for mining rules with different lengths in knowledge graph
  • Micro rule learning model for mining rules with different lengths in knowledge graph

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0075] like figure 1 As shown, this embodiment provides a differentiable rule learning model for mining rules of different lengths in a knowledge graph. The differentiable rule learning model may be called MineRAL, which is suitable for mining rules of different lengths from a knowledge graph, The technical content involved in the rules includes:

[0076] A knowledge graph is a collection of facts, a knowledge graph where, e h ,e t , r represents the head entity, the tail entity and the relationship respectively; ε is a set composed of entities; is a set of relations;

[0077] A rule in the knowledge graph consists of a rule header and a rule body, the rule header is an atomq(x,y), and the rule body is a set of atomr i (·,·); an atom is a fact that can contain variables at head and tail positions. For example, r(e h ,e t ) is an atom, where x and y are variables that can be replaced with entities;

[0078] An example of a rule in a knowledge graph is: mother(C,A)←fa...

Embodiment 2

[0136] This embodiment uses the differentiable rule learning model of the present invention to mine the chain logic rules from the knowledge graph, and describes its application, including:

[0137] (1) Training stage:

[0138] Given a knowledge graph dataset containing several facts, divide it into a training set and a validation set. The gradient descent algorithm is used to optimize the overall loss function of the differentiable rule learning model of the present invention on the training set, and when its performance on the verification set is stable, the parameters of the model are saved;

[0139] (2) Rule mining stage:

[0140] Given a query, input it into the trained differentiable rule learning model, and use the algorithm in (rule restoration) to restore the mined chain logic rules.

[0141] From the above, it can be seen that the differentiable rule learning model of the embodiment of the present invention can accurately and efficiently mine all chain logic rules ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a microgradable rule learning model for mining rules with different lengths in a knowledge graph, which comprises a player unit comprising a plurality of player modules arranged in parallel, and each player module can find queries of chained logic rules with different lengths from the knowledge graph according to the input queries of the chained logic rules with different lengths. Finding a corresponding chain logic rule from the knowledge graph; the leader unit is in communication connection with the player unit and can select a proper player module from the player unit to process a corresponding query; and the judgment unit is in communication connection with the leader unit and the player unit, can evaluate the chain logic rules found by the player unit to obtain an evaluation feedback result, and is used for the player unit and the leader unit to improve themselves according to the evaluation feedback result. The microgradable rule learning model can accurately and efficiently mine all the chain logic rules from the knowledge graph.

Description

technical field [0001] The invention relates to the field of knowledge graphs, in particular to a differentiable rule learning model for mining rules of different lengths in the knowledge graphs. Background technique [0002] Knowledge graphs consist of a large number of fact triples, which store structured human knowledge. In recent years, knowledge graphs have made great achievements in fields such as natural language processing, intelligent question answering, recommender systems, and computer vision. Commonly used knowledge graphs usually contain billions of triples, but there will still be a large number of triples missing. Due to the extremely large scale of the knowledge graph, it is expensive to manually complete the graph. Therefore, knowledge graph reasoning techniques to automatically predict missing links based on known links in knowledge graphs have attracted much attention in recent years. [0003] Common knowledge reasoning methods include rule-based method...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/36G06F16/33G06N3/04G06N3/08
CPCG06F16/367G06F16/334G06N3/08G06N3/044
Inventor 王杰张占秋陈佳俊贺华瑞
Owner UNIV OF SCI & TECH OF CHINA
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com