Knowledge graph complementing method based on topic keyword filtering

A knowledge map and keyword technology, applied in the field of knowledge map completion, can solve the problems of complex and redundant text content, inability to complete task completion, etc., and achieve the effect of solving a large amount of noisy information, improving usability, and enhancing differentiation

Inactive Publication Date: 2019-07-05
HARBIN ENG UNIV
View PDF6 Cites 21 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the problem that the complex and redundant text content of the entity description in the existing knowledge map completion method leads to the problem that a specific completion task cannot be completed in a targeted manner, and proposes a topic-based Knowledge graph completion method based on keyword filtering

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
  • Knowledge graph complementing method based on topic keyword filtering
  • Knowledge graph complementing method based on topic keyword filtering
  • Knowledge graph complementing method based on topic keyword filtering

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0047] A knowledge map completion method based on subject keyword filtering according to the present embodiment, the method includes the following steps:

[0048] Step 1: Set the knowledge map G=(E, R, T); among them, E represents the entity set of the knowledge map, R represents the relation set in the knowledge map, and T represents the set of triples to be completed,

[0049] Step 2: Set the set of incomplete triplet elements in the knowledge map G set in step 1 as the completion task set H, and the elements in H are divided into (h, r,?) and (h,? ,t) two forms; among them, the head entity h∈E, the relationship r∈R, and the tail entity t∈E;

[0050] A completion task is a collection of many tasks. Contains a lot of triples such as: (China, capital,?), (Yao Ming, place of birth,?), (Andy Lau, birthday,?) and so on. A set of many such incomplete triples is the task to be completed;

[0051] The knowledge base is filled with a large number of triples, such as: (United State...

specific Embodiment approach 2

[0089] Different from the specific embodiment 1, in this embodiment, a knowledge graph completion method based on topic keyword filtering, in the step 7, obtain the topic calculation word vector matrix D e and D t The attention score, and then select the subject according to the attention score; and calculate the word vector matrix D for the topic e and D t The process of assigning attention scores is as follows:

[0090] Step 7.1: Obtain the attention score described by the subject, and describe the attention score a by the following formula i :

[0091]

[0092]

[0093] Among them, attention(D) represents the attention scoring result of the entity description; the attention to the i-th word in the entity description text a i is the i-th word vector d i The maximum value of the cosine similarity with all row vectors in the inference matrix W(T);

[0094] Step 7.2: In the matrix W, for the document doc i , select the top m topics with the highest probability acco...

specific Embodiment approach 3

[0098] The difference from Embodiment 1 or Embodiment 2 is that in this embodiment, a knowledge map completion method based on subject keyword filtering, in step 12, the training method for adjusting parameters through training includes: stochastic gradient descent method , Adam training method, etc.

[0099] Simulation experiment data:

[0100] In each table, the data corresponding to Topic-ADRL_j is the experimental data of the method of the present invention. Experimental results show that the algorithm outperforms previous algorithms in top ten accuracy and average ranking.

[0101] Table 1 FB15K entity prediction comparison results

[0102]

[0103] Table 2 WN18 Entity Prediction Comparison Results

[0104]

[0105] Table 3 Evaluation results of relationship prediction on FB15K

[0106]

[0107] Table 4 Evaluation results of mapping relationship attributes on FB15K

[0108]

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 graph complementing method based on topic keyword filtering, and belongs to the field of knowledge graphs. Aiming at the problem that a certain specific completiontask cannot be completed in a targeted mode due to the fact that the text content described by an entity of an existing knowledge graph completion method is complex and redundant, the invention discloses a knowledge graph complementing method based on topic keyword filtering, which is characterized by integrating an attention mechanism aiming at the problem that the description information of an entity is complex and redundant, providing a topic keyword scoring function, and evaluating the description of the entity, so that the availability of an entity description text is improved, and the problem that a large amount of noise information exists in the description text is solved. In order to further reflect the semantic relation between the entity description and the triple, the semantic pertinence of the entity description is improved through the theme semantic space model. Through the text filtering method, the specific completion task can be completed in a targeted manner.

Description

technical field [0001] The invention relates to a knowledge graph completion method, in particular to a knowledge graph completion method for filtering entity description text based on subject keywords in entity descriptions. Background technique [0002] Knowledge graph technology is widely used in the field of intelligent question answering and search. At present, although the knowledge base constructed by knowledge map technology is large in scale, its completeness is still not high. Most of the entities in the graph have neither place of birth information nor nationality information, and half of the entities contain no more than 5 relationships, so it is necessary to complete the knowledge graph. Knowledge map graph methods can be divided into two categories: one is non-translational completion methods, and the other is translational completion methods. Compared with the non-translation method, the algorithm using the translation model involves fewer calculation parame...

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): G06F16/36G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/24
Inventor 印桂生张载熙王红滨
Owner HARBIN ENG UNIV
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