A method and apparatus for processing knowledge map based on semi-supervised embedded representation model

A technology of knowledge graph and embedded representation, which is applied in the field of knowledge graph processing based on semi-supervised embedded representation model, which can solve problems such as difficult to predict whether there is a connection, performance limitations, etc.

Active Publication Date: 2019-02-26
SOUTH CHINA NORMAL UNIVERSITY
View PDF4 Cites 19 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

figure 1 Among them, the entity "Tom Hanks" plays the role in the movie "Sleepless in Seattle", and the entity "MegRyan" is actually an actor in this movie. However, it is difficult to predict the relationship between "MegRyan" and "Sleepless in Seattle" according to the existing knowledge graph technology. Whether there is a connection between them, because the existing knowledge graph embedding representation methods cannot make full use of the global stru

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
  • A method and apparatus for processing knowledge map based on semi-supervised embedded representation model
  • A method and apparatus for processing knowledge map based on semi-supervised embedded representation model
  • A method and apparatus for processing knowledge map based on semi-supervised embedded representation model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] From the perspective of graph theory, the knowledge graph is essentially a graph, which can be expressed as G=(V,E), where V represents the set of vertices in the graph, and V=v 1 ,...,v n , E represents the edge set in the graph, E=e 1 ,....,e n . For the knowledge graph, each vertex v represents an entity, and each edge e represents the relationship between two entities.

[0047] The geometric meanings of the first-order proximity and the second-order proximity in the present invention are as follows:

[0048] First-order proximity: The first-order proximity describes the similarity between a pair of entities. For any pair of entities, if v i And v j There is an edge between v i And v j There is a relationship between v i And v j The first-order proximity between them is positive. Otherwise, v i And v j The first-order proximity between is 0.

[0049] According to the above definition, it is easy to know that the key to calculating the first-order proximity is to calculate...

Embodiment 2

[0114] In this embodiment, the method described in embodiment 1 is used to realize a typical application of knowledge graph-entity classification, and the objects of the application are two popular corpora of FB15K and WIN18. First, pre-process the two popular corpora to remove all entities that are not described by the associated triples. The parameters of the preprocessed corpus are shown in Table 1.

[0115] Table 1

[0116] Corpus

#Rel

#Ent

#Train

#Valid

#Test

FB15K

1336

14885

472860

50000

57800

WIN18

18

40100

140975

5000

5000

[0117] In order to form a comparison, several current most advanced methods are selected for comparison in this embodiment: including TransE, TransD, DKRL (CNN), Jointly (LSTM) and Jointly (ALSTM). The parameters used for training of all models are the parameter settings used to obtain the best performance.

[0118] The task of entity classification is a multi-label classification task aimed at predicting entity types. Almost every entity has a typ...

Embodiment 3

[0123] In this embodiment, the method described in embodiment 1 is used to realize another typical application of knowledge graph—link prediction. The target of the application is still the two popular corpora of FB15K and WIN18, and the two popular corpora are used in the same way as in embodiment 2. Two popular corpora are preprocessed, and the parameters of the two popular corpora after preprocessing are shown in Table 1.

[0124] Link prediction is a typical task of perfecting the triple (h, r, t) of the knowledge graph, where h or t is missing, that is, given (h, r) predicts t. This task places more emphasis on ranking a set of candidate entities from the knowledge graph. This embodiment uses two metrics as evaluation indicators, namely MeanRank and Hits@10, where MeanRank is the average of the correct number of entities or relationships, and Hits@10 is the ratio of the top p effective entities or relationships in the prediction. In this embodiment, p=10 is set for the enti...

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 knowledge map processing method and a device based on a semi-supervised embedded representation model, the method comprises calculating a first-order proximity and a second-order proximity of a knowledge map, calculating the supervisory loss of the first-order proximity and the supervisory loss of the second-order proximity, establishing a linear combination of the supervisory loss of the first-order proximity and the supervisory loss of the second-order proximity, optimizing the knowledge map under the condition of minimizing the linear combination, and the like. Bycalculating the first-order proximity between any two vertices in the knowledge map, and calculating the second-order proximity according to the neighborhood structure of any two vertices in the knowledge map, At the same time, the first-order proximity and the second-order proximity are considered to jointly optimize the knowledge map so as to retain the local and global structural information ofthe knowledge map, thereby overcoming the defects caused by the dependence of the knowledge map on the characteristics of each vertex and the lack of the relationship information among the vertices in the prior art. The invention is widely applied to the field of image recognition.

Description

Technical field [0001] The invention relates to the technical field of information processing, in particular to a knowledge graph processing method and device based on a semi-supervised embedded representation model. Background technique [0002] Knowledge Graph (Knowledge Graph) can not only express Internet information in a form closer to the human cognitive world, but also provides a way to better organize, manage, utilize, and integrate massive amounts of information across information sources. At present, knowledge graph technology is mainly used to support advanced applications such as automatic question answering, intelligent semantic search and recommendation systems. A typical knowledge graph describes the entities and their relationships in the physical world in symbolic form, and its basic components are usually expressed as triples (head entity, relationship, tail entity), namely (h, r, t). The entities are connected with each other through relationships, forming a n...

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
IPC IPC(8): G06K9/62G06K9/66G06N3/04G06N3/08
CPCG06N3/084G06V30/194G06N3/045G06F18/22
Inventor 朱佳赵美华郑泽涛伦家琪黄昌勤
Owner SOUTH CHINA NORMAL UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap