update
Browse files- README.md +56 -0
- mediasum.py +3 -1
- test_data.zip +3 -0
- train_data.zip +3 -0
- val_data.zip +3 -0
README.md
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
languages:
|
3 |
+
- en
|
4 |
+
multilinguality:
|
5 |
+
- monolingual
|
6 |
+
size_categories:
|
7 |
+
- 100K<n<1M
|
8 |
+
task_categories:
|
9 |
+
- conditional-text-generation
|
10 |
+
task_ids:
|
11 |
+
- summarization
|
12 |
+
---
|
13 |
+
|
14 |
+
# MediaSum dataset for summarization
|
15 |
+
|
16 |
+
Summarization dataset copied from [MediaSum: A Large-scale Media Interview Dataset for Dialogue Summarization](https://github.com/zcgzcgzcg1/MediaSumR)
|
17 |
+
|
18 |
+
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:
|
19 |
+
```python
|
20 |
+
"ccdv/mediasum": ("document", "summary")
|
21 |
+
```
|
22 |
+
|
23 |
+
# Configs
|
24 |
+
4 possibles configs:
|
25 |
+
- `roberta` will concatenate documents with "</s>" (default)
|
26 |
+
- `newline` will concatenate documents with "\n"
|
27 |
+
- `bert` will concatenate documents with "[SEP]"
|
28 |
+
- `list` will return the list of documents instead of a single string
|
29 |
+
|
30 |
+
### Data Fields
|
31 |
+
|
32 |
+
- `id`: paper id
|
33 |
+
- `document`: a string/list containing the body of a set of documents
|
34 |
+
- `summary`: a string containing the abstract of the set
|
35 |
+
|
36 |
+
### Data Splits
|
37 |
+
|
38 |
+
This dataset has 3 splits: _train_, _validation_, and _test_. \
|
39 |
+
Token counts are white space based.
|
40 |
+
|
41 |
+
| Dataset Split | Number of Instances |
|
42 |
+
| ------------- | --------------------|
|
43 |
+
| Train | 443596 |
|
44 |
+
| Validation | 10000 |
|
45 |
+
| Test | 10000 |
|
46 |
+
|
47 |
+
|
48 |
+
# Cite original article
|
49 |
+
```
|
50 |
+
@article{zhu2021mediasum,
|
51 |
+
title={MediaSum: A Large-scale Media Interview Dataset for Dialogue Summarization},
|
52 |
+
author={Zhu, Chenguang and Liu, Yang and Mei, Jie and Zeng, Michael},
|
53 |
+
journal={arXiv preprint arXiv:2103.06410},
|
54 |
+
year={2021}
|
55 |
+
}
|
56 |
+
```
|
mediasum.py
CHANGED
@@ -21,7 +21,7 @@ _CITATION = """\
|
|
21 |
}
|
22 |
"""
|
23 |
_ABSTRACT = "summary"
|
24 |
-
_ARTICLE = "
|
25 |
|
26 |
class MediaSumSummarizationConfig(datasets.BuilderConfig):
|
27 |
"""BuilderConfig for MediaSumSummarization."""
|
@@ -117,7 +117,9 @@ class MediaSumSummarizationDataset(datasets.GeneratorBasedBuilder):
|
|
117 |
'summary': str,
|
118 |
'document': List[str],
|
119 |
"""
|
|
|
120 |
document = data["utt"]
|
|
|
121 |
if self.config.name != "list":
|
122 |
document = join_.join(document)
|
123 |
summary = data["summary"]
|
|
|
21 |
}
|
22 |
"""
|
23 |
_ABSTRACT = "summary"
|
24 |
+
_ARTICLE = "document"
|
25 |
|
26 |
class MediaSumSummarizationConfig(datasets.BuilderConfig):
|
27 |
"""BuilderConfig for MediaSumSummarization."""
|
|
|
117 |
'summary': str,
|
118 |
'document': List[str],
|
119 |
"""
|
120 |
+
|
121 |
document = data["utt"]
|
122 |
+
|
123 |
if self.config.name != "list":
|
124 |
document = join_.join(document)
|
125 |
summary = data["summary"]
|
test_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:760b1efde09738c70fdb0449dbb9a4fa07f1d474cb790ddf9446650052df89d7
|
3 |
+
size 33913771
|
train_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5d29532f5f9d84b33a9007ebad450898540742edff55d5d058fa47e4c975a3b7
|
3 |
+
size 1442774847
|
val_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8f466af07260fd2fa72aff5b065ac355e9bf677931b3e84c99764e6ebf2c9215
|
3 |
+
size 34058060
|