16. Knowledge Graph

16.1 Overview

Paper

• 【Great】Knowledge Graphs - 2020

  Article: 史上最全《知识图谱》2020综述论文，18位作者, 130页pdf

• Learning Entity and Relation Embeddings for Knowledge Graph Completion - THU2015

• COMET: Commonsense Transformers for Automatic Knowledge Graph Construction - Allen2019

  Chinese: ACL 2019 | AI2等提出自动知识图谱构建模型COMET，接近人类表现

• How much is a Triple? Estimating the Cost of Knowledge Graph Creation - Germany2018

  Chinese: 67 亿美金搞个图，创建知识图谱的成本有多高你知道吗？

• https://github.com/liuhuanyong/KnowledgeGraphSlides

  中文知识图谱计算会议CCKS报告合集，涵盖从2013年至2018年，共48篇，从中可以看出从谷歌2012年推出知识图谱以来，中国学术界及工业界这6年来知识图谱的主流思想变迁。

• Distant supervision for relation extraction without labeled data - Stanford2009

  远程监督 = 监督学习 + Bootstrapping

  Article: 关系抽取之远程监督算法 - 2019

Data

• https://www.ownthink.com/

  思知 智能时代：史上最大规模1.4亿中文知识图谱开源下载

  Code:https://github.com/ownthink/KnowledgeGraphData

Code

• https://github.com/thunlp/KB2E (C++)

• https://github.com/wuxiyu/transE (Python)

• https://github.com/liuhuanyong/CausalityEventExtraction

  基于因果关系知识库的因果事件图谱实验项目，本项目罗列了因果显式表达的几种模式，基于这种模式和大规模语料，再经过融合等操作，可形成因果事件图谱。

• https://github.com/liuhuanyong/HyponymyExtraction

  基于知识概念体系，百科知识库，以及在线搜索结构化方式的词语上下位抽取与可视化展示

• https://github.com/liuhuanyong/EventTriplesExtraction

  基于依存句法与语义角色标注的事件三元组抽取，可用于文本理解如文档主题链，事件线等应用

Competition

瑞金医院MMC人工智能辅助构建知识图谱大赛

• https://github.com/zhpmatrix/tianchi-ruijin (Keras)

  baseline, BiLSTM+CRF, 只做了第一阶段实体识别，作为NER来做，第二阶段是关系抽取，其实可以End-to-End一起做

2019百度信息抽取比赛

抽取满足约束的SPO三元组知识, http://lic2019.ccf.org.cn/kg

• https://github.com/bojone/kg-2019 (Keras)

  Rank 7 基于CNN + Attenton + 自行设计的标注结构的信息抽取模型

• https://github.com/zhengyima/kg-baseline-pytorch (PyTorch)

  使用Pytorch实现苏神的模型，F1在dev集可达到0.75，联合关系抽取，Joint Relation Extraction.

• https://github.com/wangpeiyi9979/IE-Bert-CNN (PyTorch)

  BERT + CNN

Article

• 知识图谱从哪里来：实体关系抽取的现状与未来 - 2019

16.2 Representation Learning

• https://github.com/chenbjin/RepresentationLearning

  知识表示相关学习算法

• https://github.com/liuhuanyong/ProductKnowledgeGraph

  基于京东网站的商品上下级概念、商品品牌之间关系和商品描述维度等知识库，可以支持商品属性库构建、商品销售问答、品牌物品生产等知识查询服务，也可用于情感分析等下游应用。

16.3 Entity

• https://github.com/hendrydong/Chinese-Event-Extraction

  ?

Article

• 如何为实体抽取描述性短语 - 2020

16.4 Relation

Practice

• https://github.com/thunlp/JointNRE (Tensorflow)

  Joint Neural Relation Extraction with Text and KGs

  Paper: 1

• https://github.com/shiningliang/CCKS2019-IPRE (Tensorflow)

  CCKS2019-人物关系抽取

• https://github.com/mengxiaoxu/entity_relation_extraction (Java)

  基于依存分析的实体关系抽取简单实现，即抽取事实三元组

• https://github.com/lawlietAi/relation-classification-via-attention-model (PyTorch)

  code of Relation Classification via Multi-Level Attention CNNs

• https://github.com/hadyelsahar/CNN-RelationExtraction (Tensorflow)

  CNN for relation extraction between two given entities

• https://github.com/lemonhu/open-entity-relation-extraction

  Knowledge triples extraction and knowledge base construction based on dependency syntax for open domain text.

Paper

• 0. More Data, More Relations, More Context and More Openness: A Review and Outlook for Relation Extraction - THU2020

  Article: 【清华大学-腾讯】关系提取综述

• 1. Neural Knowledge Acquisition via Mutual Attention between Knowledge Graph and Text - THU2018
• 2. Chinese Relation Extraction with Multi-Grained Information and External Linguistic Knowledge - THU2019

  Code: https://github.com/thunlp/Chinese_NRE (PyTorch)

Article

• 关系提取简述 - 2019

• 【论文】Awesome Relation Classification Paper - 2019

  PART I

  PART II

• 【论文】Awesome Relation Extraction Paper - 2019

  PART I

  PART II

16.5 End-to-end

Practice

• https://github.com/gswycf/Joint-Extraction-of-Entities-and-Relations-Based-on-a-Novel-Tagging-Scheme (PyTorch)

  Joint Extraction of Entities and Relations Based on cnn+rnn

  Paper: paper1, 2

• https://github.com/zsctju/triplets-extraction (Keras)

  Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme

  Paper: paper1

• https://github.com/WindChimeRan/pytorch_multi_head_selection_re (PyTorch)

  reproduce "Joint entity recognition and relation extraction as a multi-head selection problem" for Chinese IE

  Paper: paper3

• https://github.com/sanmusunrise/NPNs

  Paper: paper4

Paper

• 1. Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme - CAS2017
• 2. Deep Active Learning for Named Entity Recognition - UTEXAS2018
• 3. Joint entity recognition and relation extraction as a multi-head selection problem - Belgium2018
• 4. Nugget Proposal Networks for Chinese Event Detection - CAS2018
• 5. ointly Multiple Events Extraction via Attention-based Graph Information Aggregation - BIT2018