Author Disambiguation using Knowledge Graph Embeddings with Literals

Related tags

Deep Learningand-kge
Overview

Author Name Disambiguation with Knowledge Graph Embeddings using Literals

This is the repository for the master thesis project on Knowledge Graph Embeddings for Author Name Disambiguation presented by Cristian Santini at Digital Humanities and Digital Knowledge - University of Bologna, with the collaboration of Information Service Engineering - FIZ Karlsruhe, in a.y. 2020/2021.

Datasets

This repository contains notebooks and scripts used for a research on Author Name Disambiguation using Knowledge Graph Embeddings (KGEs) with literals. Due to the unavailability of an established benchmark for evaluating our approach, we extracted two Knowledge Graphs (KGs) from the following publicly available resources: 1) a triplestore available on Zenodo [1] covering information about the journal Scientometrics and modelled according to the OpenCitations Data Model and 2) a publicly available benchmark for author disambiguation available at this link by AMiner.
The Knowledge Graphs extracted are available on Zenodo as OpenCitations-782K [2] and AMiner-534K [3]. Each dataset is organized as a collection of RDF triples stored in TSV format. Literal triples are stored separately in order to train multimodal Knowledge Graph Embedding models.
Each dataset contains a JSON file called and_eval.json which contains a list of publications in the scholarly KGs labelled for evaluating AND algorithms. For the evaluation, while for AMiner-534K the set of publications was already manually annotated by a team of experts, for OC-782K we used the ORCID iDs associated with the authors in the triplestore in order to create an evaluation dataset.

PyKEEN extension

The pykeen-extension directory contains extension files compatible with PyKEEN (Release: v1.4.0.). In this directory we implemented some extensions of the LiteralE model [4] which allow to train multimodal knowledge graph embeddings by also using textual information contained in entity descriptions. Details about the models and on how to install the extension files are available in this README.md file.
The extended library provides an implementation of the following models:

  • DistMultText: an extension of the DistMult model [5] for training KGEs by using entity descriptions attached to entities.
  • ComplExText: an extension of the ComplEx model [6] which allows to train KGEs by using information coming from short text descriptions attached to entities.
  • DistMult_gate_text: an extension of the DistMult model which allows to train KGEs by using information coming from short text descriptions and numeric value associated with entities in KGs.
    Entity descriptions are encoded by using SPECTER [7], a BERT language model for scientific documents.

Code

Scripts used in our research are available in the src directory. The disambiguation.py file in the src/disambiguation folder contains the functions that we developed for carrying author name disambiguation by using knowledge graph embeddings. More specifically it contains:

  • the do_blocking() function, which is used to preliminarily group the authors in the KG into different sub-sets by means of their last name and first initial,
  • the cluster_KGEs() function, which takes as input the output of the do_blocking function and disambiguates the authors by means of Knowledge Graph Embeddings and Hierarchical Agglomerative Clustering.
  • the evaluation functions that we used in our experiments. The src folder also contains the various scripts used for extracting the scholarly KGs from the original sources and creating an evaluation dataset for AND.

Results

Knowledge Graph Embedding Evaluation

For evaluating the quality of our KGE models in representing the components of the studied KGs, OpenCitations-782K and AMiner-534K, we used entity prediction, one of the most common KG-completion tasks. In our experiments, we compared three architectures:

  • A DistMult model trained with only structural triples, i.e. triples connecting just two entities.
  • A DistMultText model which was trained by using titles of scholarly resources, i.e. journals and publications, along with structural triples.
  • A DistMult_gate_text model which was trained using titles and publication dates of scholarly resources in order to leverage the representations of the entities associated with them.
    Hyper-parameters were obtained by doing hyper-parameter optimization with PyKEEN. Details about the configuration files are available in the kge-evaluation folder.

The following table shows the results of our experiments for OC-782K.

Model MR MRR [email protected] [email protected] [email protected] [email protected]
DistMult 59901 0.3570 0.3157 0.3812 0.402 0.4267
DistMultText 60495 0.3568 0.3158 0.3809 0.4013 0.4252
DistMult_gate_text 61812 0.3534 0.3130 0.3767 0.3971 0.4218

The following table shows the results of our experiments for AMiner-534K.

Model MR MRR [email protected] [email protected] [email protected] [email protected]
DistMult 3585 0.3285 0.1938 0.3996 0.4911 0.5940
DistMultText 3474 0.3443 0.2139 0.4123 0.5014 0.6019
DistMult_gate_text 3560 0.3452 0.2163 0.4123 0.5009 0.6028

Author Name Disambiguation

We compared our architecture for Author Name Disambiguation (AND) for KGs with a simple Rule-based method inspired by Caron and Van Eck [8] on OC-782K and with other state-of-the-art graph embedding models on the AMiner benchmark (results taken from [9]). The results are reported below.

Model Precision Recall F1
Caron & Van Eck [8] 84.66 50.20 63.03
DistMult 91.71 67.11 77.50
DistMultText 89.63 66.98 76.67
DistMult_gate_text 82.76 67.59 74.40

Model Precision Recall F1
Zhang and Al Hasan [10] 70.63 59.53 62.81
Zhang et Al. [9] 77.96 63.03 67.79
DistMult 78.36 59.68 63.36
DistMultText 77.24 61.21 64.18
DistMult_gate_text 77.62 59.91 63.07

References

[1] Massari, Arcangelo. (2021). Bibliographic dataset based on Scientometrics, containing provenance information compliant with the OpenCitations Data Model and non disambigued authors (1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5151264

[2] Santini, Cristian, Alam, Mehwish, Gesese, Genet Asefa, Peroni, Silvio, Gangemi, Aldo, & Sack, Harald. (2021). OC-782K: Knowledge Graph of "Scientometrics" modelled according to the OpenCitations Data Model [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5675787

[3] Santini, Cristian, Alam, Mehwish, Gesese. Genet Asefa, Peroni, Silvio, Gangemi, Aldo, & Sack, Harald. (2021). AMiner-534K: Knowledge Graph of AMiner benchmark for Author Name Disambiguation [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5675801

[4] Kristiadi A., Khan M.A., Lukovnikov D., Lehmann J., Fischer A. (2019) Incorporating Literals into Knowledge Graph Embeddings. In: Ghidini C. et al. (eds) The Semantic Web – ISWC 2019. ISWC 2019. Lecture Notes in Computer Science, vol 11778. Springer, Cham. https://doi.org/10.1007/978-3-030-30793-6_20.

[5] Yang, B., Yih, W., He, X., Gao, J., & Deng, L. (2015). Embedding Entities and Relations for Learning and Inference in Knowledge Bases. ArXiv:1412.6575 [Cs]. http://arxiv.org/abs/1412.6575

[6] Trouillon, T., Welbl, J., Riedel, S., Gaussier, É., & Bouchard, G. (2016). Complex Embeddings for Simple Link Prediction. ArXiv:1606.06357 [Cs, Stat]. http://arxiv.org/abs/1606.06357

[7] Cohan, A., Feldman, S., Beltagy, I., Downey, D., & Weld, D. S. (2020). SPECTER: Document-level Representation Learning using Citation-informed Transformers. ArXiv:2004.07180 [Cs]. http://arxiv.org/abs/2004.07180

[8] Caron, E., & van Eck, N.-J. (2014). Large scale author name disambiguation using rule-based scoring and clustering: International conference on science and technology indicators. Proceedings of the Science and Technology Indicators Conference 2014, 79–86. http://sti2014.cwts.nl

[9] Zhang, Y., Zhang, F., Yao, P., & Tang, J. (2018). Name Disambiguation in AMiner: Clustering, Maintenance, and Human in the Loop. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 1002–1011. https://doi.org/10.1145/3219819.3219859

[10] Zhang, B., & Al Hasan, M. (2017). Name Disambiguation in Anonymized Graphs using Network Embedding. Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, 1239–1248. https://doi.org/10.1145/3132847.3132873

Acknowledgments

The software and data here available are the result of a master thesis carried in collaboration between the FICLIT department of the University of Bologna and the research department FIZ - Information Service Engineering (ISE) of the Karlsruhe Institute of Technology (KIT). The thesis has been supervised by Prof. Aldo Gangemi and Prof. Silvio Peroni from the University of Bologna, and Prof. Harald Sack, Dr. Mehwish Alam and Genet Asefa Gesese from FIZ-ISE.

Kaggle | 9th place single model solution for TGS Salt Identification Challenge

UNet for segmenting salt deposits from seismic images with PyTorch. General We, tugstugi and xuyuan, have participated in the Kaggle competition TGS S

Erdene-Ochir Tuguldur 276 Dec 20, 2022
Automatic self-diagnosis program (python required)Automatic self-diagnosis program (python required)

auto-self-checker 자동으로 자가진단 해주는 프로그램(python 필요) 중요 이 프로그램이 실행될때에는 절대로 마우스포인터를 움직이거나 키보드를 건드리면 안된다(화면인식, 마우스포인터로 직접 클릭) 사용법 프로그램을 구동할 폴더 내의 cmd창에서 pip

1 Dec 30, 2021
TVNet: Temporal Voting Network for Action Localization

TVNet: Temporal Voting Network for Action Localization This repo holds the codes of paper: "TVNet: Temporal Voting Network for Action Localization". P

hywang 5 Jul 26, 2022
Unofficial PyTorch implementation of TokenLearner by Google AI

tokenlearner-pytorch Unofficial PyTorch implementation of TokenLearner by Ryoo et al. from Google AI (abs, pdf) Installation You can install TokenLear

Rishabh Anand 46 Dec 20, 2022
Code and model benchmarks for "SEVIR : A Storm Event Imagery Dataset for Deep Learning Applications in Radar and Satellite Meteorology"

NeurIPS 2020 SEVIR Code for paper: SEVIR : A Storm Event Imagery Dataset for Deep Learning Applications in Radar and Satellite Meteorology Requirement

USAF - MIT Artificial Intelligence Accelerator 46 Dec 15, 2022
QTool: A Low-bit Quantization Toolbox for Deep Neural Networks in Computer Vision

This project provides abundant choices of quantization strategies (such as the quantization algorithms, training schedules and empirical tricks) for quantizing the deep neural networks into low-bit c

Monash Green AI Lab 51 Dec 10, 2022
Project ArXiv Citation Network

Project ArXiv Citation Network Overview This project involved the analysis of the ArXiv citation network. Usage The complete code of this project is i

Dennis Núñez-Fernández 5 Oct 20, 2022
Neural Surface Maps

Neural Surface Maps Official implementation of Neural Surface Maps - Luca Morreale, Noam Aigerman, Vladimir Kim, Niloy J. Mitra [Paper] [Project Page]

Luca Morreale 49 Dec 13, 2022
DGL-TreeSearch and the Gurobi-MWIS interface

Independent Set Benchmarking Suite This repository contains the code for our maximum independent set benchmarking suite as well as our implementations

Maximilian Böther 19 Nov 22, 2022
Github project for Attention-guided Temporal Coherent Video Object Matting.

Attention-guided Temporal Coherent Video Object Matting This is the Github project for our paper Attention-guided Temporal Coherent Video Object Matti

71 Dec 19, 2022
Model serving at scale

Run inference at scale Cortex is an open source platform for large-scale machine learning inference workloads. Workloads Realtime APIs - respond to pr

Cortex Labs 7.9k Jan 06, 2023
VR Viewport Pose Model for Quantifying and Exploiting Frame Correlations

This repository contains the introduction to the collected VRViewportPose dataset and the code for the IEEE INFOCOM 2022 paper: "VR Viewport Pose Model for Quantifying and Exploiting Frame Correlatio

0 Aug 10, 2022
Gans-in-action - Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks

GANs in Action by Jakub Langr and Vladimir Bok List of available code: Chapter 2: Colab, Notebook Chapter 3: Notebook Chapter 4: Notebook Chapter 6: C

GANs in Action 914 Dec 21, 2022
ViewFormer: NeRF-free Neural Rendering from Few Images Using Transformers

ViewFormer: NeRF-free Neural Rendering from Few Images Using Transformers Official implementation of ViewFormer. ViewFormer is a NeRF-free neural rend

Jonáš Kulhánek 169 Dec 30, 2022
ICRA 2021 "Towards Precise and Efficient Image Guided Depth Completion"

PENet: Precise and Efficient Depth Completion This repo is the PyTorch implementation of our paper to appear in ICRA2021 on "Towards Precise and Effic

232 Dec 25, 2022
Code release for Local Light Field Fusion at SIGGRAPH 2019

Local Light Field Fusion Project | Video | Paper Tensorflow implementation for novel view synthesis from sparse input images. Local Light Field Fusion

1.1k Dec 27, 2022
Pyramid Scene Parsing Network, CVPR2017.

Pyramid Scene Parsing Network by Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, Jiaya Jia, details are in project page. Introduction This

Hengshuang Zhao 1.5k Jan 05, 2023
DuBE: Duple-balanced Ensemble Learning from Skewed Data

DuBE: Duple-balanced Ensemble Learning from Skewed Data "Towards Inter-class and Intra-class Imbalance in Class-imbalanced Learning" (IEEE ICDE 2022 S

6 Nov 12, 2022
PyTorch implementations for our SIGGRAPH 2021 paper: Editable Free-viewpoint Video Using a Layered Neural Representation.

st-nerf We provide PyTorch implementations for our paper: Editable Free-viewpoint Video Using a Layered Neural Representation SIGGRAPH 2021 Jiakai Zha

Diplodocus 258 Jan 02, 2023
Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data

Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data This is the official PyTorch implementation of the SeCo paper: @articl

ElementAI 101 Dec 12, 2022