Knowledge Graph Paper Review System for Academic Papers
Discover how a Knowledge Graph-based paper review system facilitates information extraction, knowledge graph construction, and academic paper analysis. Explore structural information extraction, two-stage refinement of knowledge graphs, and the application in a paper review system.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
A Knowledge Graph Based Paper Review System Structure Information Extraction & Knowledge Graph Construction for Academic Paper ZHONGYE WANG, 517030910353 YICHEN XIE, 517030910355 XINYU ZHAN, 517030910358
Content 1. Introduction 2. Structural Information Extraction 3. Two-stage Refinement of Knowledge Graph 4. Conclusions
What is Structure Information? Title Title Abstract PDF Paper Abstract Section Section 1 Subsection Subsection ... PDF/Image Text Sequence Document Tree Knowledge Graph Pool Machine Understanding Good Machine Understanding
Our Framework Structural Information Extraction & Primitive KG Construction Kowledge Graph Refinement & Application: Paper Review System Fine Tune Refinement Model Information Extraction Review Model PDF Paper Score: 98 PDFX Embedding GNN Primitive KG Refined KG
Our Framework Structural Information Extraction & Primitive KG Construction Kowledge Graph Refinement & Application: Paper Review System Fine Tune Refinement Model Information Extraction Review Model PDF Paper Score: 98 PDFX Embedding GNN Primitive KG Refined KG
Our Framework Structural Information Extraction & Primitive KG Construction Kowledge Graph Refinement & Application: Paper Review System Fine Tune Refinement Model Information Extraction Review Model PDF Paper Score: 98 PDFX Embedding GNN Primitive KG Refined KG
Our Framework Structural Information Extraction & Primitive KG Construction Kowledge Graph Refinement & Application: Paper Review System Fine Tune Refinement Model Information Extraction Review Model PDF Paper Score: 98 PDFX Embedding GNN Primitive KG Refined KG
Content 1. Introduction 2. Structural Information Extraction 3. Two-stage Refinement of Knowledge Graph 4. Conclusions
Information Extraction of PDF Title Abstract Author Keywords Text Body Section Heading Footnote Footnote
Construct XML from PDF Construct geometric model to determine spatial organism Identify different logical units Label XML file to indicate logical structure
Label XML file with CRF 3y 1y 2y 4y Abstract Author Title Keywords Influence Influence Influence ... 3 x 1x 2 x 4 x exp( ( , , , ,...)) F x y y y = 1 2 Pr( | ) i i i y x i i ( ) Z x
Primitive Knowledge Graph Entities Elements from the XML output Relations Structure Relation Reference Relation
Primitive Knowledge Graph Example Parent: Lead: Caption: Refer:
Content 1. Introduction 2. Structural Information Extraction 3. Two-stage Refinement of Knowledge Graph 4. Conclusions
Motivation Fine Tune Refinement Model Review Model Score: 98 Embedding GNN Primitive KG Refined KG There may be missing relations, and we need graph completion. The structure relations and reference relations are too naive. A relation may be composed of multiple sub-relations.
Stage: Embed KG with the Refinement Model
Stage: Embed KG with the Refinement Model
Stage: Embed KG with the Refinement Model
Stage: Embed KG with the Refinement Model Scoring Function
Stage: Embed KG with the Refinement Model Scoring Function
Stage: Scoring with Sub-Relations For entities A and B, Relation R Sub-relation R1 Sub-relation R2 Sub-relation R3 1rv 2rv 3rv subrelation scoring functions ( , , ) ( , , ) ( , , ) f v v v f v v v f v v v a b 1r a b 2r a b 3r aggregate( , S ( , )) (3 r , ) = ( ( , )) S r a b ( )) S a b (1 r ( , )) (2 r , S a b a b = = Pr( ( , )) ( ( ( , ))) r a b S r a b Pr( ( , )) ( ( ( , ))) r a b S r a b i i
Stage: Results Score Function: - Existing Trans-E model - MLP trained from scratch Aggregate Method: - Maximum - Summation - Softmax Adjusted Summation Optimal: MLP + Adjusted Sum
Stage: Results Figure to the right shows distributions of subrelations. The degeneration case (all subrelations are the same) doesn t happen.
Stage: Finetune with the Review Model Unsupervised Learning Sub-Relations Lacks INFORMATION! Introduce More INFORMATION!!
Stage: Finetune with the Review Model
Stage: Finetune with the Review Model
Stage: Finetune with the Review Model
Stage: Finetune with the Review Model
Stage: Finetune with the Review Model
Content 1. Introduction 2. Structural Information Extraction 3. Two-stage Refinement of Knowledge Graph 4. Conclusions
Conclusion Fine Tune Refinement Model Information Extraction Review Model PDF Paper Score: 98 PDFX Embedding GNN Primitive KG Refined KG Pros: Refined KGs carry more logical information of the paper. A review model for analyzing paper performance. Cons: Sub-relations lack interpretabiltiy.
