Hugging Face's logo

Datasets:
linxy
/
ICEWS05_15

Tasks:
Graph Machine Learning
Languages:
English
Size:
1M<n<10M
ArXiv:
License:
Dataset card Files Community
ICEWS05_15 / README.md
linxy's picture
linxy
Update README.md
17a1a7a verified
preview code
|
raw
history blame contribute delete
10.9 kB
metadata
license: apache-2.0
task_categories:
  - graph-ml
language:
  - en
size_categories:
  - 1M<n<10M

TL;DR: The datasets for the temporal knowledge graph reasoning task.

[Github] [OpenReview] [arXiv]

  • Built over ICEWS and GDELT, which are widely used benchmarks in TKGC.
  • First introduced in paper "TFLEX: Temporal Feature-Logic Embedding Framework for Complex Reasoning over Temporal Knowledge Graph"
  • Please refer to the original paper for more details.

See also: [ICEWS14] [GDELT]

🔬 Usage 

>>> dataset = load_dataset("linxy/ICEWS05_15", "all")
>>> len(dataset["train"]) + len(dataset["validation"]) + len(dataset["test"])
4651939
>>> dataset["train"][0]
{'query_name': 'Pe',
 'definition': 'def Pe(e1, r1, t1): return Pe(e1, r1, t1)',
 'query': [3751, 125, 1330],
 'answer': [10136],
 'easy_answer': [],
 'args': ['e1', 'r1', 't1']}
>>> dataset["test"][0]
{'query_name': 'Pe2',
 'definition': 'def Pe2(e1, r1, t1, r2, t2): return Pe(Pe(e1, r1, t1), r2, t2)',
 'query': [7262, 425, 3943, 144, 2619],
 'answer': [2473, 5870],
 'easy_answer': [5870],
 'args': ['e1', 'r1', 't1', 'r2', 't2']}

'args' is the argument list of the query function, where name starting with 'e' is entity, and 'r' for relation, 't' for timestamp.

assert len(query) == len(args)

In order to decode query ids into text, we should use a vocabulary (i.e. entity2idx, relation2idx and timestamp2idx). Therefore, we use the code below to load meta info which contains the vocabulary:

>>> dataset = load_dataset("linxy/ICEWS05_15", "meta")
>>> meta_info = dataset_meta["train"][0]
>>> meta_info
{'dataset': 'ICEWS05_15',
 'entity_count': 10488,
 'relation_count': 251,
 'timestamp_count': 4017,
 'valid_triples_count': 46275,
 'test_triples_count': 46092,
 'train_triples_count': 368962,
 'triple_count': 461329,
 'query_meta': {'query_name': [...], 'queries_count': [...], 'avg_answers_count': [...], ...},
 'entity2idx': {'name': [...], 'id': [...]},
 'relation2idx': {'name': [...], 'id': [...]},
 'timestamp2idx': {'name': [...], 'id': [...]},

Since the ids in the vocabulary are already sorted, we directly decode to access the name text:

>>> query
[3751, 125, 1330]
>>> args
['e1', 'r1', 't1']
>>> for idx, arg_type in zip(query, args):
        if arg_type.startswith('e') or arg_type.startswith('s') or arg_type.startswith('o'):  # s, o, e1, e2, ...
            print(idx, meta_info['entity2idx']['name'][idx])
        elif arg_type.startswith('r'):  # r, r1, r2, ...
            print(idx, meta_info['relation2idx']['name'][idx])
        elif arg_type.startswith('t'):  # t, t1, t2, ...
            print(idx, meta_info['timestamp2idx']['name'][idx])

Besides, we also provide query-type-specific subparts.

>>> dataset = load_dataset("linxy/ICEWS05_15", "e2i")
>>> some_datasets = [load_dataset("linxy/ICEWS05_15", query_name) for query_name in meta_info['query_meta']['query_name']]

Help yourself!

👈 🔎 Dataset statistics: queries_count
query ICEWS14 ICEWS05_15 GDELT
train valid test train valid test train valid test
Pe 66783 8837 8848 344042 45829 45644 1115102 273842 273432
Pe2 72826 3482 4037 368962 10000 10000 2215309 10000 10000
Pe3 72826 3492 4083 368962 10000 10000 2215309 10000 10000
e2i 72826 3305 3655 368962 10000 10000 2215309 10000 10000
e3i 72826 2966 3023 368962 10000 10000 2215309 10000 10000
Pt 42690 7331 7419 142771 28795 28752 687326 199780 199419
aPt 13234 4411 4411 68262 10000 10000 221530 10000 10000
bPt 13234 4411 4411 68262 10000 10000 221530 10000 10000
Pe_Pt 7282 3385 3638 36896 10000 10000 221530 10000 10000
Pt_sPe_Pt 13234 5541 6293 68262 10000 10000 221530 10000 10000
Pt_oPe_Pt 13234 5480 6242 68262 10000 10000 221530 10000 10000
t2i 72826 5112 6631 368962 10000 10000 2215309 10000 10000
t3i 72826 3094 3296 368962 10000 10000 2215309 10000 10000
e2i_N 7282 2949 2975 36896 10000 10000 221530 10000 10000
e3i_N 7282 2913 2914 36896 10000 10000 221530 10000 10000
Pe_e2i_Pe_NPe 7282 2968 3012 36896 10000 10000 221530 10000 10000
e2i_PeN 7282 2971 3031 36896 10000 10000 221530 10000 10000
e2i_NPe 7282 3061 3192 36896 10000 10000 221530 10000 10000
t2i_N 7282 3135 3328 36896 10000 10000 221530 10000 10000
t3i_N 7282 2924 2944 36896 10000 10000 221530 10000 10000
Pe_t2i_PtPe_NPt 7282 3031 3127 36896 10000 10000 221530 10000 10000
t2i_PtN 7282 3300 3609 36896 10000 10000 221530 10000 10000
t2i_NPt 7282 4873 5464 36896 10000 10000 221530 10000 10000
e2u - 2913 2913 - 10000 10000 - 10000 10000
Pe_e2u - 2913 2913 - 10000 10000 - 10000 10000
t2u - 2913 2913 - 10000 10000 - 10000 10000
Pe_t2u - 2913 2913 - 10000 10000 - 10000 10000
t2i_Pe - 2913 2913 - 10000 10000 - 10000 10000
Pe_t2i - 2913 2913 - 10000 10000 - 10000 10000
e2i_Pe - 2913 2913 - 10000 10000 - 10000 10000
Pe_e2i - 2913 2913 - 10000 10000 - 10000 10000
between 7282 2913 2913 36896 10000 10000 221530 10000 10000
Pe_aPt 7282 4134 4733 68262 10000 10000 221530 10000 10000
Pe_bPt 7282 3970 4565 36896 10000 10000 221530 10000 10000
Pt_sPe 7282 4976 5608 36896 10000 10000 221530 10000 10000
Pt_oPe 7282 3321 3621 36896 10000 10000 221530 10000 10000
Pt_se2i 7282 3226 3466 36896 10000 10000 221530 10000 10000
Pt_oe2i 7282 3236 3485 36896 10000 10000 221530 10000 10000
Pe_at2i 7282 4607 5338 36896 10000 10000 221530 10000 10000
Pe_bt2i 7282 4583 5386 36896 10000 10000 221530 10000 10000
👈 🔎 Dataset statistics: avg_answers_count
query ICEWS14 ICEWS05_15 GDELT
train valid test train valid test train valid test
Pe 1.09 1.01 1.01 1.07 1.01 1.01 2.07 1.21 1.21
Pe2 1.03 2.19 2.23 1.02 2.15 2.19 2.61 6.51 6.13
Pe3 1.04 2.25 2.29 1.02 2.18 2.21 5.11 10.86 10.70
e2i 1.02 2.76 2.84 1.01 2.36 2.52 1.05 2.30 2.32
e3i 1.00 1.57 1.59 1.00 1.26 1.26 1.00 1.20 1.35
Pt 1.71 1.22 1.21 2.58 1.61 1.60 3.36 1.66 1.66
aPt 177.99 176.09 175.89 2022.16 2003.85 1998.71 156.48 155.38 153.41
bPt 181.20 179.88 179.26 1929.98 1923.75 1919.83 160.38 159.29 157.42
Pe_Pt 1.58 7.90 8.62 2.84 18.11 20.63 26.56 42.54 41.33
Pt_sPe_Pt 1.79 7.26 7.47 2.49 13.51 10.86 4.92 14.13 12.80
Pt_oPe_Pt 1.75 7.27 7.48 2.55 13.01 14.34 4.62 14.47 12.90
t2i 1.19 6.29 6.38 3.07 29.45 25.61 1.97 8.98 7.76
t3i 1.01 2.88 3.14 1.08 10.03 10.22 1.06 3.79 3.52
e2i_N 1.02 2.10 2.14 1.01 2.05 2.08 2.04 4.66 4.58
e3i_N 1.00 1.00 1.00 1.00 1.00 1.00 1.02 1.19 1.37
Pe_e2i_Pe_NPe 1.04 2.21 2.25 1.02 2.16 2.19 3.67 8.54 8.12
e2i_PeN 1.04 2.22 2.26 1.02 2.17 2.21 3.67 8.66 8.36
e2i_NPe 1.18 3.03 3.11 1.12 2.87 2.99 4.00 8.15 7.81
t2i_N 1.15 3.31 3.44 1.21 4.06 4.20 2.91 8.78 7.56
t3i_N 1.00 1.02 1.03 1.01 1.02 1.02 1.15 3.19 3.20
Pe_t2i_PtPe_NPt 1.08 2.59 2.70 1.08 2.47 2.62 4.10 12.02 11.37
t2i_PtN 1.41 5.22 5.47 1.70 8.10 8.11 4.56 12.56 11.32
t2i_NPt 8.14 25.96 26.23 66.99 154.01 147.34 17.58 35.60 32.22
e2u 0.00 3.12 3.17 0.00 2.38 2.40 0.00 5.04 5.41
Pe_e2u 0.00 2.38 2.44 0.00 1.24 1.25 0.00 9.39 10.78
t2u 0.00 4.35 4.53 0.00 5.57 5.92 0.00 9.70 10.51
Pe_t2u 0.00 2.72 2.83 0.00 1.24 1.28 0.00 9.90 11.27
t2i_Pe 0.00 1.03 1.03 0.00 1.01 1.02 0.00 1.34 1.44
Pe_t2i 0.00 1.14 1.16 0.00 1.07 1.08 0.00 2.01 2.20
e2i_Pe 0.00 1.00 1.00 0.00 1.00 1.00 0.00 1.07 1.10
Pe_e2i 0.00 2.18 2.24 0.00 1.32 1.33 0.00 5.08 5.49
between 122.61 120.94 120.27 1407.87 1410.39 1404.76 214.16 210.99 207.85
Pe_aPt 4.67 16.73 16.50 18.68 43.80 46.23 49.31 66.21 68.88
Pe_bPt 4.53 17.07 16.80 18.70 45.81 48.23 67.67 84.79 83.00
Pt_sPe 8.65 28.86 29.22 71.51 162.36 155.46 27.55 45.83 43.73
Pt_oPe 1.41 5.23 5.46 1.68 8.36 8.21 3.84 11.31 10.06
Pt_se2i 1.31 5.72 6.19 1.37 9.00 9.30 2.76 8.72 7.66
Pt_oe2i 1.32 6.51 7.00 1.44 10.49 10.89 2.55 8.17 7.27
Pe_at2i 7.26 22.63 21.98 30.40 60.03 53.18 88.77 101.60 101.88
Pe_bt2i 7.27 21.92 21.23 30.31 61.59 64.98 88.80 100.64 100.67

✉️ Contact

🤝 Citation

Please condiser citing this paper if you use the code or data from our work. Thanks a lot :)

(Xueyuan et al., 2023 preferred, instead of Lin et al., 2023)

@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}
}

TFLEX is released under the Apache License 2.0 license.

(back to top)