Datasets:
add readme
Browse files
README.md
ADDED
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
languages:
|
3 |
+
- en
|
4 |
+
multilinguality:
|
5 |
+
- monolingual
|
6 |
+
size_categories:
|
7 |
+
- 10K<n<100K
|
8 |
+
task_categories:
|
9 |
+
- conditional-text-generation
|
10 |
+
task_ids:
|
11 |
+
- summarization
|
12 |
+
---
|
13 |
+
|
14 |
+
# GovReport dataset for summarization
|
15 |
+
|
16 |
+
Dataset for summarization of long documents.\
|
17 |
+
Adapted from this [repo](https://github.com/luyang-huang96/LongDocSum).\
|
18 |
+
Note that original data are pre-tokenized so this dataset returns " ".join(text).\
|
19 |
+
This dataset is compatible with the [`run_summarization.py`](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization) script from Transformers if you add this line to the `summarization_name_mapping` variable:
|
20 |
+
```python
|
21 |
+
"ccdv/govreport-summarization": ("report", "summary")
|
22 |
+
```
|
23 |
+
|
24 |
+
### Data Fields
|
25 |
+
|
26 |
+
- `id`: paper id
|
27 |
+
- `report`: a string containing the body of the report
|
28 |
+
- `summary`: a string containing the summary of the report
|
29 |
+
|
30 |
+
### Data Splits
|
31 |
+
|
32 |
+
This dataset has 3 splits: _train_, _validation_, and _test_. \
|
33 |
+
Token counts with a RoBERTa tokenizer.
|
34 |
+
|
35 |
+
| Dataset Split | Number of Instances | Avg. tokens |
|
36 |
+
| ------------- | --------------------|:----------------------|
|
37 |
+
| Train | 17,517 | < 9,000 / < 500 |
|
38 |
+
| Validation | 973 | < 9,000 / < 500 |
|
39 |
+
| Test | 973 | < 9,000 / < 500 |
|
40 |
+
|
41 |
+
|
42 |
+
# Cite original article
|
43 |
+
```
|
44 |
+
@misc{huang2021efficient,
|
45 |
+
title={Efficient Attentions for Long Document Summarization},
|
46 |
+
author={Luyang Huang and Shuyang Cao and Nikolaus Parulian and Heng Ji and Lu Wang},
|
47 |
+
year={2021},
|
48 |
+
eprint={2104.02112},
|
49 |
+
archivePrefix={arXiv},
|
50 |
+
primaryClass={cs.CL}
|
51 |
+
}
|
52 |
+
```
|
53 |
+
|