Skip to content
Permalink
Browse files

Added new Chinese and Korean reading comprehension datasets

  • Loading branch information
sebastianruder committed Feb 24, 2020
1 parent fdaf509 commit 726d01ab169f2e117e8eb9aa7f9f0575726e680a
Showing with 66 additions and 11 deletions.
  1. +5 −0 README.md
  2. +0 −11 chinese/chinese.md
  3. +45 −0 chinese/question_answering.md
  4. +16 −0 korean/question_answering.md
@@ -59,6 +59,7 @@

- [Entity linking](chinese/chinese.md#entity-linking)
- [Chinese word segmentation](chinese/chinese_word_segmentation.md)
- [Question answering](chinese/question_answering.md)

### French

@@ -76,6 +77,10 @@

- [Question Answering](portuguese/question_answering.md)

### Korean

- [Question Answering](korean/question_answering.md)

### Nepali

- [Machine Translation](nepali/nepali.md#machine-translation)
@@ -16,14 +16,3 @@ See [here](../english/entity_linking.md) for more information about the task.
| Tsai & Roth (2016) | 83.6 | [Cross-lingual wikification using multilingual embeddings](http://cogcomp.org/papers/TsaiRo16b.pdf) | |

[Go back to the README](../README.md)

## reading comprehension

### Dureader Datasets
See [here](https://ai.baidu.com/broad/introduction?dataset=dureader) to see the introduction.

#### Baidu DuReader Dataset
See [here](https://ai.baidu.com/broad/download?dataset=dureader) to download the Dataset.

#### Disambiguation-Only Models
See [here](https://ai.baidu.com/broad/leaderboard?dataset=dureader) to see the leaderboard.
@@ -0,0 +1,45 @@
# Question answering

Question answering is the task of answering a question.

### Table of contents

- [Reading comprehension](#reading-comprehension)
- [CMRC2018](#cmrc-2018)
- [DRCD](#drcd)
- [DuReader](#dureader)

## Reading comprehension

### CMRC 2018

The [Chinese Machine Reading Comprehension (CMRC 2018)](https://www.aclweb.org/anthology/D19-1600/) is a SQuAD-like
reading comprehension dataset that consists of 20,000 questions annotated on Wikipedia paragraphs by human experts. The
dataset can be downloaded [here](https://github.com/ymcui/cmrc2018). Below we show the F1 and EM scores both on the
test set and the challenge set.

| Model | Test F1 | Test EM | Challenge F1 | Challenge EM | Paper |
| ------------- | :-----:| :-----:| --- |
| Human performance | 97.9 | 92.4 | 95.2 | 90.4 | [A Span-Extraction Dataset for Chinese Machine Reading Comprehension](https://www.aclweb.org/anthology/D19-1600/) |
| Dual BERT (w / SQuAD; Cui et al., 2019) | 90.2 | 73.6 | 55.2 | 27.8 | [Cross-Lingual Machine Reading Comprehension](https://www.aclweb.org/anthology/D19-1169/) |
| Dual BERT (Cui et al., 2019) | 88.1 | 70.4 | 47.9 | 23.8 | [Cross-Lingual Machine Reading Comprehension](https://www.aclweb.org/anthology/D19-1169/) |

### DRCD

The [Delta Reading Comprehension Dataset (DRCD)](https://arxiv.org/abs/1806.00920) is a SQuAD-like reading
comprehension dataset that contains 30,000+ questions on 10,014 paragraphs from 2,108 Wikipedia articles. The dataset
can be downloaded [here](https://github.com/DRCKnowledgeTeam/DRCD).

| Model | F1 | EM | Paper |
| ------------- | :-----:| :-----:| --- |
| Human performance | 93.3 | 80.4 | [DRCD: a Chinese Machine Reading Comprehension Dataset](https://arxiv.org/abs/1806.00920) |
| Dual BERT (w / SQuAD; Cui et al., 2019) | 91.6 | 85.4 | [Cross-Lingual Machine Reading Comprehension](https://www.aclweb.org/anthology/D19-1169/) |
| Dual BERT (Cui et al., 2019) | 90.3 | 83.7 | [Cross-Lingual Machine Reading Comprehension](https://www.aclweb.org/anthology/D19-1169/) |

### DuReader

[DuReader](https://www.aclweb.org/anthology/W18-2605/) is a large-scale reading comprehension dataset that is based on
the logs of Baidu Search and contains 200k questions, 420k answers, and 1M documents. For more information, refer to
[its website](https://ai.baidu.com/broad/introduction?dataset=dureader) to see the introduction. You can download the
dataset [here](https://ai.baidu.com/broad/download?dataset=dureader). The best models can be view on the
[public leaderboard](https://ai.baidu.com/broad/leaderboard?dataset=dureader).
@@ -0,0 +1,16 @@
# Question answering

Question answering is the task of answering a question.

### Table of contents

- [Reading comprehension](#reading-comprehension)
- [KorQuAD](#korquad)

## Reading comprehension

### KorQuAD

The [Korean Question Answering Dataset (KorQuAD)](https://arxiv.org/abs/1909.07005) is a large-scale reading comprehension
dataset in the style of SQuAD that consists of 70,000+ human-generated question answer pairs on Wikipedia articles. The
data and public leaderboard are available [here](https://korquad.github.io/).

0 comments on commit 726d01a

Please sign in to comment.
You can’t perform that action at this time.