linxy/ICEWS14
收藏资源简介:
该数据集用于时间知识图谱推理任务。它基于ICEWS和GDELT构建,这两个是时间知识图谱补全(TKGC)中广泛使用的基准。该数据集首次在论文《TFLEX: 用于时间知识图谱复杂推理的时间特征-逻辑嵌入框架》中提出。数据集中包含多种查询类型(如Pe、Pe2、e2i等),每条样本包括查询名称、定义、查询ID列表、答案ID列表、简易答案列表以及参数列表(参数中e开头表示实体,r开头表示关系,t开头表示时间戳)。此外,还提供了词汇表(实体、关系、时间戳到ID的映射)以及各查询类型的统计信息(样本数量、平均答案数等)。数据许可证为Apache-2.0。
This dataset is used for temporal knowledge graph reasoning tasks. It is built upon ICEWS and GDELT, which are widely used benchmarks in temporal knowledge graph completion (TKGC). The dataset was first introduced in the paper "TFLEX: Temporal Feature-Logic Embedding Framework for Complex Reasoning over Temporal Knowledge Graph". It contains multiple query types (e.g., Pe, Pe2, e2i, etc.), and each sample includes a query name, a definition, a list of query IDs, a list of answer IDs, a list of easy answers, and a list of arguments (where arguments starting with e denote entities, r denotes relations, and t denotes timestamps). Additionally, the dataset provides vocabularies (mapping from entities, relations, and timestamps to IDs) as well as statistics for each query type (e.g., number of samples, average number of answers). The data license is Apache-2.0.
数据集概述
基本信息
- 许可证: Apache-2.0
- 任务类别: 图机器学习(Graph-ML)
- 语言: 英语(en)
- 数据集大小: 1M<n<10M
数据集内容
- 构建基础: 基于ICEWS和GDELT,这两个数据集在时序知识图谱补全(TKGC)领域广泛使用。
- 首次引入: 在论文"TFLEX: Temporal Feature-Logic Embedding Framework for Complex Reasoning over Temporal Knowledge Graph"中首次介绍。
数据集结构
-
数据加载示例: python
dataset = load_dataset("linxy/ICEWS14", "all") len(dataset["train"]) + len(dataset["validation"]) + len(dataset["test"]) 1088769
-
数据样本示例: python {query_name: Pe_aPt, definition: def Pe_aPt(e1, r1, e2, r2, e3): return Pe(e1, r1, after(Pt(e2, r2, e3))), query: [6291, 372, 5683, 283, 5264], answer: [1077], easy_answer: [], args: [e1, r1, e2, r2, e3]}
-
元数据信息: python {dataset: ICEWS14, entity_count: 7128, relation_count: 230, timestamp_count: 365, valid_triples_count: 8941, test_triples_count: 8963, train_triples_count: 72826, triple_count: 90730, query_meta: {query_name: [...], queries_count: [...], avg_answers_count: [...], ...}, entity2idx: {name: [...], id: [...]}, relation2idx: {name: [...], id: [...]}, timestamp2idx: {name: [...], id: [...]},
数据集统计
- 查询计数统计: 提供了不同查询类型在训练、验证和测试集中的数量。
- 平均答案计数统计: 提供了不同查询类型的平均答案数量。
数据集使用
- 数据加载: 支持加载完整数据集、元数据以及特定查询类型的子部分。
- 数据解码: 使用元数据中的词汇表(entity2idx, relation2idx, timestamp2idx)将查询ID解码为文本。
联系方式
- 联系人: Lin Xueyuan
- 邮箱: linxy59@mail2.sysu.edu.cn
引用信息
- 论文: TFLEX: Temporal Feature-Logic Embedding Framework for Complex Reasoning over Temporal Knowledge Graph
- 作者: Lin Xueyuan et al.
- 会议: Thirty-seventh Conference on Neural Information Processing Systems
- 年份: 2023
- 引用格式: bibtex @inproceedings{ xueyuan2023tflex, title={TFLEX: Temporal Feature-Logic Embedding Framework for Complex Reasoning over Temporal Knowledge Graph}, author={Lin Xueyuan and Haihong E and Chengjin Xu and Gengxian Zhou and Haoran Luo and Tianyi Hu and Fenglong Su and Ningyuan Li and Mingzhi Sun}, booktitle={Thirty-seventh Conference on Neural Information Processing Systems}, year={2023}, url={https://openreview.net/forum?id=oaGdsgB18L} }




