File size: 2,462 Bytes
20ed7b1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
# Analogy Test Dataset
This dataset contains five different word analogy datasets. Each contains jsonline files for validation and test, in which each line consists of following dictionary,
```
{"stem": ["raphael", "painter"],
"answer": 2,
"choice": [["andersen", "plato"],
["reading", "berkshire"],
["marx", "philosopher"],
["tolstoi", "edison"]]}
```
where `stem` is the query word pair, `choice` has word pair candidates,
and `answer` indicates the index of correct candidate which starts from `0`. Data statistics are summarized as below.
| Dataset | Size (valid/test) | Num of choice | Num of relation group | Original Reference |
|---------|------------------:|--------------:|----------------------:|:--------------------------------------------------------------------------:|
| sat | 37/337 | 5 | 2 | [Turney (2005)](https://arxiv.org/pdf/cs/0508053.pdf) |
| u2 | 24/228 | 5,4,3 | 9 | [EnglishForEveryone](https://englishforeveryone.org/Topics/Analogies.html) |
| u4 | 48/432 | 5,4,3 | 5 | [EnglishForEveryone](https://englishforeveryone.org/Topics/Analogies.html) |
| google | 50/500 | 4 | 2 | [Mikolov et al., (2013)](https://www.aclweb.org/anthology/N13-1090.pdf) |
| bats | 199/1799 | 4 | 3 | [Gladkova et al., (2016)](https://www.aclweb.org/anthology/N18-2017.pdf) |
All data is lowercased except Google dataset. Please read [our paper](https://arxiv.org/abs/2105.04949) for more information about the dataset and cite it if you use the dataset:
```
@inproceedings{ushio-etal-2021-bert-is,
title ={{BERT} is to {NLP} what {A}lex{N}et is to {CV}: {C}an {P}re-{T}rained {L}anguage {M}odels {I}dentify {A}nalogies?},
author={Ushio, Asahi and
Espinosa-Anke, Luis and
Schockaert, Steven and
Camacho-Collados, Jose},
booktitle={Proceedings of the {ACL}-{IJCNLP} 2021 Main Conference},
year={2021},
publisher={Association for Computational Linguistics}
}
```
### LICENSE
The LICENSE of all the resources are under [CC-BY-NC-4.0](./LICENSE). Thus, they are freely available for academic purpose or individual research, but restricted for commercial use.
|