Giters

yuqianghan / KHG-Papers

Paper list for knowledge hypergraph

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Paper List for Knowledge Hypergraph

This is a collection of research papers on knowledge hypergraph, a.k.a., n-ary relational knowledge base (KB), higher-arity KB, multi-fold relational data, hyper-relational knowledge graph (KG).

Overview

Fact Definition

  • tuple definition (N-ary relation)
  • role-entity definition
  • triplet-qualifier definition (hyper-relation)

Research Papers

tuple definition

  • m-TransH: On the Representation and Embedding of Knowledge Bases Beyond Binary Relations. Jianfeng Wen, Jianxin Li, Yongyi Mao, Shini Chen, and Ruichong Zhang. IJCAI 2016. paper code

    m-TransH is directly extended from TransH in KG (tuple definition adopted).

  • RAE: Scalable Instance Reconstruction in Knowledge Bases via Relatedness Affiliated Embedding. Ruichong Zhang, Junpeng Li, Jiajie Mei, and Yongyi Mao. WWW 2018. paper code

    RAE further considers relatedness between entities using MLP, based on m-TransH (tuple definition adopted).

  • GETD: Generalizing Tensor Decomposition for N-ary Relational Knowledge Bases. Yu Liu, Quanming Yao, and Yong Li. WWW 2020. paper code

    GETD extends TuckER in KG with Tensor Ring decomposition utilized, which applies for n-ary relational KBs with single arity (tuple definition adopted).

  • HypE: Knowledge Hypergraphs: Prediction Beyond Binary Relations. Bahare Fatemi, Perouz Taslakian, David Vazquez, and David Poole. IJCAI 2020. paper code

    HypE uses convolution filters to obtain entity embeddings, which are used to calculate score by multilinear product (tuple definition adopted).

  • BoxE: A Box Embedding Model for Knowledge Base Completion. Ralph Abboud, ˙Ismail ˙Ilkan Ceylan, Thomas Lukasiewicz, and Tommaso Salvatori. NeurIPS 2020. paper code

    BoxE utilizes hyper-rectangles to obtain plausibility score for n-ary relational facts (tuple definition adopted).

  • G-MPNN: Neural Message Passing for Multi-Relational Ordered and Recursive Hypergraphs. Naganand Yadati. NeurIPS 2020. paper code

    G-MPNN unifies message passing for multi-relational ordered hypergraphs, which provides representation for link prediction in knowledge hypergraph (tuple definition adopted).

  • Searching to Sparsify Tensor Decomposition for N-ary Relational Data. Shiming Di, Quanming Yao, and Lei Chen. WWW 2021. paper code

    Leveraging AutoML for link prediction in n-ary relational KBs (tuple definition adopted).

  • ReAIE: Knowledge Hypergraph Embedding Meets Relational Algebra. Bahare Fatemi, Perouz Taslakian, David Vazquez, and David Poole. arxiv 2021. paper code

    ReAIE introduces the window concept to model the interaction of different elements of the embeddings, which are used to calculate dot scoring function (tuple definition adopted).

role-entity

  • NaLP: Link Prediction on N-ary Relational Data. Saiping Guan, Xiaolong Jin, Yuanzhuo Wang, and Xueqi Cheng. WWW 2019. paper code

    NaLP concatenates role and value embeddings, which are fed into convolution layer and fully connected layer to measure the score (role-value definition adopted).

  • RAM: Role-Aware Modeling for N-ary Relational Knowledge Bases. Yu Liu, Quanming Yao, and Yong Li. WWW 2021. paper code

    RAM introduces latent space and role-aware pattern matrix for multilinear scoring function (role-value definition adopted).

  • Link Prediction on N-ary Relational Data Based on Relatedness Evaluation. Saiping Guan, Xiaolong Jin, Jiafeng Guo, Yuanzhuo Wang, and Xueqi Cheng. TKDE 2021. paper code

    The journal extension of NaLP in WWW'19.

triplet-qualifier

  • HINGE: Beyond Triplets: Hyper-Relational Knowledge Graph Embedding for Link Prediction. Paolo Rosso, Dingqi Yang, and Philippe Cudré-Mauroux. WWW 2020. paper code

    HINGE learns from triplet and attribute-value pairs with convolutional layer, which are both fed into fully connected layer for final score (triplet-qualifier definition adopted).

  • NeuInfer: Knowledge Inference on N-ary Facts. Saiping Guan, Xiaolong Jin, Jiafeng Guo, Yuanzhuo Wang, and Xueqi Cheng. ACL 2020. paper code

    NeuInfer learns from triplet and attribute-value pairs with fully connected layer, which are both fed into fully connected layer for final score (triplet-qualifier definition adopted).

  • StarE: Message Passing for Hyper-Relational Knowledge Graphs. Mikhail Galkin, Priyansh Trivedi, Gaurav Maheshwari, Ricardo Usbeck, and Jens Lehmann. EMNLP 2020. paper code

    StarE utilizes CompGCN to update embeddings, which are further utilized for score by Transformer (triplet-qualifier definition adopted).

Public Dataset:

About

Paper list for knowledge hypergraph