Datasets:
Tasks:
Text Generation
Modalities:
Text
Sub-tasks:
language-modeling
Languages:
Korean
Size:
10K - 100K
ArXiv:
Tags:
question-generation
License:
license: cc-by-4.0 | |
pretty_name: KorQuAD for question generation | |
language: ko | |
multilinguality: monolingual | |
size_categories: 10K<n<100K | |
source_datasets: squad_es | |
task_categories: question-generation | |
task_ids: question-generation | |
# Dataset Card for "lmqg/qg_korquad" | |
## Dataset Description | |
- **Repository:** [https://github.com/asahi417/lm-question-generation](https://github.com/asahi417/lm-question-generation) | |
- **Paper:** [TBA](TBA) | |
- **Point of Contact:** [Asahi Ushio](http://asahiushio.com/) | |
### Dataset Summary | |
This is a subset of [QG-Bench](https://github.com/asahi417/lm-question-generation/blob/master/QG_BENCH.md#datasets), a unified question generation benchmark proposed in | |
["Generative Language Models for Paragraph-Level Question Generation: A Unified Benchmark and Evaluation, EMNLP 2022 main conference"](paper_link). | |
This is a modified version of [KorQuAD](https://huggingface.co./datasets/squad_kor_v1) for question generation (QG) task. | |
Since the original dataset only contains training/validation set, we manually sample test set from training set, which | |
has no overlap in terms of the paragraph with the training set. | |
### Supported Tasks and Leaderboards | |
* `question-generation`: The dataset is assumed to be used to train a model for question generation. | |
Success on this task is typically measured by achieving a high BLEU4/METEOR/ROUGE-L/BERTScore/MoverScore (see our paper for more in detail). | |
### Languages | |
Korean (ko) | |
## Dataset Structure | |
An example of 'train' looks as follows. | |
``` | |
{ | |
"question": "ν¨μν΄μνμ΄ μ£Όλͺ©νλ νꡬλ?", | |
"paragraph": "λ³νμ λν μ΄ν΄μ λ¬μ¬λ μμ°κ³Όνμ μμ΄μ μΌλ°μ μΈ μ£Όμ μ΄λ©°, λ―Έμ λΆνμ λ³νλ₯Ό νꡬνλ κ°λ ₯ν λꡬλ‘μ λ°μ λμλ€. ν¨μλ λ³ννλ μμ λ¬μ¬ν¨μ μμ΄μ μ€μΆμ μΈ κ°λ μΌλ‘μ¨ λ μ€λ₯΄κ² λλ€. μ€μμ μ€λ³μλ‘ κ΅¬μ±λ ν¨μμ μλ°ν νκ΅¬κ° μ€ν΄μνμ΄λΌλ λΆμΌλ‘ μλ €μ§κ² λμκ³ , 볡μμμ λν μ΄μ κ°μ νꡬλΆμΌλ 볡μν΄μνμ΄λΌκ³ νλ€. ν¨μν΄μνμ ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬμ μ£Όλͺ©νλ€. ν¨μν΄μνμ λ§μ μμ©λΆμΌ μ€ νλκ° μμμνμ΄λ€. λ§μ λ¬Έμ λ€μ΄ μμ°μ€λ½κ² μκ³Ό κ·Έ μμ λ³νμ¨μ κ΄κ³λ‘ κ·μ°©λκ³ , μ΄λ¬ν λ¬Έμ λ€μ΄ λ―ΈλΆλ°©μ μμΌλ‘ λ€λ£¨μ΄μ§λ€. μμ°μ λ§μ νμλ€μ΄ λμνκ³λ‘ κΈ°μ λ μ μλ€. νΌλ μ΄λ‘ μ μ΄λ¬ν μμΈ‘ λΆκ°λ₯ν νμμ νꡬνλ λ° μλΉν κΈ°μ¬λ₯Ό νλ€.", | |
"answer": "ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬ", | |
"sentence": "ν¨μν΄μνμ ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬ μ μ£Όλͺ©νλ€.", | |
"paragraph_sentence": 'λ³νμ λν μ΄ν΄μ λ¬μ¬λ μμ°κ³Όνμ μμ΄μ μΌλ°μ μΈ μ£Όμ μ΄λ©°, λ―Έμ λΆνμ λ³νλ₯Ό νꡬνλ κ°λ ₯ν λꡬλ‘μ λ°μ λμλ€. ν¨μλ λ³ννλ μμ λ¬μ¬ν¨μ μμ΄μ μ€μΆμ μΈ κ°λ μΌλ‘μ¨ λ μ€λ₯΄κ² λλ€. μ€μμ μ€λ³μλ‘ κ΅¬μ±λ ν¨μμ μλ°ν νκ΅¬κ° μ€ν΄μνμ΄λΌλ λΆμΌλ‘ μλ €μ§κ² λμκ³ , 볡μμμ λν μ΄μ κ°μ νꡬ λΆμΌλ 볡μν΄μνμ΄λΌκ³ νλ€. <hl> ν¨μν΄μνμ ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬ μ μ£Όλͺ©νλ€. <hl> ν¨μν΄μνμ λ§μ μμ©λΆμΌ μ€ νλκ° μμμνμ΄λ€. λ§μ λ¬Έμ λ€μ΄ μμ°μ€λ½κ² μκ³Ό κ·Έ μμ λ³νμ¨μ κ΄κ³λ‘ κ·μ°©λκ³ , μ΄λ¬ν λ¬Έμ λ€μ΄ λ―ΈλΆλ°©μ μμΌλ‘ λ€λ£¨μ΄μ§λ€. μμ°μ λ§μ νμλ€μ΄ λμνκ³λ‘ κΈ°μ λ μ μλ€. νΌλ μ΄λ‘ μ μ΄λ¬ν μμΈ‘ λΆκ°λ₯ν νμμ νꡬνλ λ° μλΉν κΈ°μ¬λ₯Ό νλ€.', | |
"paragraph_answer": 'λ³νμ λν μ΄ν΄μ λ¬μ¬λ μμ°κ³Όνμ μμ΄μ μΌλ°μ μΈ μ£Όμ μ΄λ©°, λ―Έμ λΆνμ λ³νλ₯Ό νꡬνλ κ°λ ₯ν λꡬλ‘μ λ°μ λμλ€. ν¨μλ λ³ννλ μμ λ¬μ¬ν¨μ μμ΄μ μ€μΆμ μΈ κ°λ μΌλ‘μ¨ λ μ€λ₯΄κ² λλ€. μ€μμ μ€λ³μλ‘ κ΅¬μ±λ ν¨μμ μλ°ν νκ΅¬κ° μ€ν΄μνμ΄λΌλ λΆμΌλ‘ μλ €μ§κ² λμκ³ , 볡μμμ λν μ΄μ κ°μ νꡬ λΆμΌλ 볡μν΄μνμ΄λΌκ³ νλ€. ν¨μν΄μνμ <hl> ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬ <hl>μ μ£Όλͺ©νλ€. ν¨μν΄μνμ λ§μ μμ©λΆμΌ μ€ νλκ° μμμνμ΄λ€. λ§μ λ¬Έμ λ€μ΄ μμ°μ€λ½κ² μκ³Ό κ·Έ μμ λ³νμ¨μ κ΄κ³λ‘ κ·μ°©λκ³ , μ΄λ¬ν λ¬Έμ λ€μ΄ λ―ΈλΆλ°©μ μμΌλ‘ λ€λ£¨μ΄μ§λ€. μμ°μ λ§μ νμλ€μ΄ λμνκ³λ‘ κΈ°μ λ μ μλ€. νΌλ μ΄λ‘ μ μ΄λ¬ν μμΈ‘ λΆκ°λ₯ν νμμ νꡬνλ λ° μλΉν κΈ°μ¬λ₯Ό νλ€.', | |
"sentence_answer": "ν¨μν΄μνμ <hl> ν¨μμ 곡κ°(νΉν 무νμ°¨μ)μ νꡬ <hl> μ μ£Όλͺ©νλ€." | |
} | |
``` | |
The data fields are the same among all splits. | |
- `question`: a `string` feature. | |
- `paragraph`: a `string` feature. | |
- `answer`: a `string` feature. | |
- `sentence`: a `string` feature. | |
- `paragraph_answer`: a `string` feature, which is same as the paragraph but the answer is highlighted by a special token `<hl>`. | |
- `paragraph_sentence`: a `string` feature, which is same as the paragraph but a sentence containing the answer is highlighted by a special token `<hl>`. | |
- `sentence_answer`: a `string` feature, which is same as the sentence but the answer is highlighted by a special token `<hl>`. | |
Each of `paragraph_answer`, `paragraph_sentence`, and `sentence_answer` feature is assumed to be used to train a question generation model, | |
but with different information. The `paragraph_answer` and `sentence_answer` features are for answer-aware question generation and | |
`paragraph_sentence` feature is for sentence-aware question generation. | |
## Data Splits | |
|train|validation|test | | |
|----:|---------:|----:| | |
|54556| 5766 |5766 | | |
## Citation Information | |
``` | |
@inproceedings{ushio-etal-2022-generative, | |
title = "{G}enerative {L}anguage {M}odels for {P}aragraph-{L}evel {Q}uestion {G}eneration: {A} {U}nified {B}enchmark and {E}valuation", | |
author = "Ushio, Asahi and | |
Alva-Manchego, Fernando and | |
Camacho-Collados, Jose", | |
booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing", | |
month = dec, | |
year = "2022", | |
address = "Abu Dhabi, U.A.E.", | |
publisher = "Association for Computational Linguistics", | |
} | |
``` |