Added new Chinese and Korean reading comprehension datasets

README.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)

chinese/chinese.md

@@ -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.

chinese/question_answering.md

@@ -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).

korean/question_answering.md

@@ -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/).