AHE alignment hyperplane-based text knowledge embedding method

A hyper-planar, text-based technology, applied in the field of knowledge graphs, can solve problems such as inaccuracy and inflexibility, and achieve the effect of expanding the basic model and improving the effect of completion

Active Publication Date: 2021-03-23
FUZHOU UNIVERSITY
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of this, the purpose of the present invention is to provide a text knowledge embedding method based on the AHE aligned hyperplane, which solves the inaccuracy and inflexibility introduced by the traditional text joint learning model to the text description, and further effectively improves the knowledge map completion. Effect

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
  • AHE alignment hyperplane-based text knowledge embedding method
  • AHE alignment hyperplane-based text knowledge embedding method
  • AHE alignment hyperplane-based text knowledge embedding method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0086] In this embodiment, in order to verify the effectiveness of the present invention, knowledge map completion is performed on the data sets FB-Text and WN9-Text, and each data set is divided into a training set and a test set.

[0087] For each triplet in the test set, use all entities in the knowledge graph to replace its tail entity or head entity (not at the same time), to generate a new triplet, and use the new score f r (s,o) to rate it. After sorting these scores in descending order, the ranking order of the original triples in a query is obtained. The mean rank (MeanRank, MR) and the proportion of test triplets with rank less than or equal to N (Hits@N, N=1, 3, 10) and the mean reciprocal rank (Mean Reciprocal Rank, MRR) are used as evaluation indicators. Lower MR and higher Hits@N, MRR all mean better performance of the model.

[0088] In this embodiment, Adam is used as the optimization algorithm, and the grid search method is used to find the most suitable hyp...

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 relates to an AHE alignment hyperplane-based text knowledge embedding method, which comprises the following steps: S1, performing word vector pre-training on a data set to obtain accurate representation of text description; and S2, aligning the text vector and the internal vector into a unified hidden layer dimension by adopting an AHE alignment hyperplane strategy, respectively carrying out text hyperplane projection on the head entity and the tail entity to obtain an interaction enhanced representation vector, and applying the interaction enhanced representation vector to the knowledge base basic model. The problem that a traditional text joint learning model introduces inaccuracy and inflexibility to text description is solved, and the knowledge graph completion effect isfurther effectively improved.

Description

technical field [0001] The invention relates to the field of knowledge graphs, in particular to a text knowledge embedding method based on AHE aligned hyperplanes. Background technique [0002] Existing methods have achieved some success in exploiting textual descriptions, but there are still some problems. in DKRL ] In , each entity is associated with a structural representation vector and a description representation vector, but the final joint model uses a simple first-order constraint to learn two types of vectors at the same time, which cannot effectively capture the correlation between text and triples . The SSP method of Han et al. improved this by establishing a head-tail description-specific semantic hyperplane to project the internal structure score, and trade-off between the internal structure score and the projected score by the λ hyperparameter, which is in a certain It can be accurately captured to a certain extent, but its performance depends on the semanti...

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): G06N5/02G06F16/36
CPCG06N5/022G06N5/027G06F16/367
Inventor 汪璟玢张旺
Owner FUZHOU 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