Datasets:
updated load dataset file
Browse files
scientific_lay_summarization.py
CHANGED
|
@@ -39,6 +39,14 @@ _DESCRIPTION = """
|
|
| 39 |
This repository contains the PLOS and eLife datasets, introduced in the EMNLP 2022 paper "[Making Science Simple: Corpora for the Lay Summarisation of Scientific Literature
|
| 40 |
](https://arxiv.org/abs/2210.09932)".
|
| 41 |
Each dataset contains full biomedical research articles paired with expert-written lay summaries (i.e., non-technical summaries). PLOS articles are derived from various journals published by [the Public Library of Science (PLOS)](https://plos.org/), whereas eLife articles are derived from the [eLife](https://elifesciences.org/) journal. More details/anlaysis on the content of each dataset are provided in the paper.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
"""
|
| 43 |
|
| 44 |
_DOCUMENT = "article"
|
|
@@ -59,7 +67,7 @@ class ScientificLaySummarisationConfig(datasets.BuilderConfig):
|
|
| 59 |
filename: filename of different configs for the dataset.
|
| 60 |
**kwargs: keyword arguments forwarded to super.
|
| 61 |
"""
|
| 62 |
-
super(ScientificLaySummarisationConfig, self).__init__(version=datasets.Version("1.0"), **kwargs)
|
| 63 |
self.filename = filename
|
| 64 |
|
| 65 |
|
|
@@ -78,7 +86,7 @@ class ScientificLaySummarisation(datasets.GeneratorBasedBuilder):
|
|
| 78 |
{
|
| 79 |
_DOCUMENT: datasets.Value("string"),
|
| 80 |
_SUMMARY: datasets.Value("string"),
|
| 81 |
-
"
|
| 82 |
"keywords": datasets.Value("string"),
|
| 83 |
"year": datasets.Value("string"),
|
| 84 |
"title": datasets.Value("string"),
|
|
@@ -117,19 +125,22 @@ class ScientificLaySummarisation(datasets.GeneratorBasedBuilder):
|
|
| 117 |
# "id": str, # unique identifier
|
| 118 |
# "year": int, # year of publication
|
| 119 |
# "title": str, # title
|
| 120 |
-
# "sections": List[List[str]], # main text, divided in to sections
|
| 121 |
# "headings" List[str], # headings of each section
|
| 122 |
-
# "abstract": List[str], # abstract
|
| 123 |
-
# "summary": List[str], # lay summary
|
| 124 |
# "keywords": List[str] # keywords/topic of article
|
| 125 |
|
| 126 |
d = json.loads(line)
|
| 127 |
-
|
|
|
|
|
|
|
|
|
|
| 128 |
|
| 129 |
yield d["id"], {
|
| 130 |
-
_DOCUMENT: "\n".join([
|
| 131 |
_SUMMARY: summary,
|
| 132 |
-
"section_headings": "\n".join(d["headings"]),
|
| 133 |
"keywords": "\n".join(d["keywords"]),
|
| 134 |
"year": d["year"],
|
| 135 |
"title": d["title"]
|
|
|
|
| 39 |
This repository contains the PLOS and eLife datasets, introduced in the EMNLP 2022 paper "[Making Science Simple: Corpora for the Lay Summarisation of Scientific Literature
|
| 40 |
](https://arxiv.org/abs/2210.09932)".
|
| 41 |
Each dataset contains full biomedical research articles paired with expert-written lay summaries (i.e., non-technical summaries). PLOS articles are derived from various journals published by [the Public Library of Science (PLOS)](https://plos.org/), whereas eLife articles are derived from the [eLife](https://elifesciences.org/) journal. More details/anlaysis on the content of each dataset are provided in the paper.
|
| 42 |
+
|
| 43 |
+
Both "elife" and "plos" have 6 features:
|
| 44 |
+
- "article": the body of the document (including the abstract), sections seperated by "/n".
|
| 45 |
+
- "section_headings": the title of each section, seperated by "/n".
|
| 46 |
+
- "keywords": keywords describing the topic of the article, seperated by "/n".
|
| 47 |
+
- "title" : the title of the article.
|
| 48 |
+
- "year" : the year the article was published.
|
| 49 |
+
- "summary": the lay summary of the document.
|
| 50 |
"""
|
| 51 |
|
| 52 |
_DOCUMENT = "article"
|
|
|
|
| 67 |
filename: filename of different configs for the dataset.
|
| 68 |
**kwargs: keyword arguments forwarded to super.
|
| 69 |
"""
|
| 70 |
+
super(ScientificLaySummarisationConfig, self).__init__(version=datasets.Version("1.0.0"), **kwargs)
|
| 71 |
self.filename = filename
|
| 72 |
|
| 73 |
|
|
|
|
| 86 |
{
|
| 87 |
_DOCUMENT: datasets.Value("string"),
|
| 88 |
_SUMMARY: datasets.Value("string"),
|
| 89 |
+
"section_headings": datasets.Value("string"),
|
| 90 |
"keywords": datasets.Value("string"),
|
| 91 |
"year": datasets.Value("string"),
|
| 92 |
"title": datasets.Value("string"),
|
|
|
|
| 125 |
# "id": str, # unique identifier
|
| 126 |
# "year": int, # year of publication
|
| 127 |
# "title": str, # title
|
| 128 |
+
# "sections": List[List[str]], # main text, divided in to sections/sentences
|
| 129 |
# "headings" List[str], # headings of each section
|
| 130 |
+
# "abstract": List[str], # abstract, in sentences
|
| 131 |
+
# "summary": List[str], # lay summary, in sentences
|
| 132 |
# "keywords": List[str] # keywords/topic of article
|
| 133 |
|
| 134 |
d = json.loads(line)
|
| 135 |
+
|
| 136 |
+
sections = [" ".join(s).strip() for s in d["sections"]]
|
| 137 |
+
abstract = " ".join(d['abstract']).strip()
|
| 138 |
+
summary = " ".join(d["summary"]).strip()
|
| 139 |
|
| 140 |
yield d["id"], {
|
| 141 |
+
_DOCUMENT: "\n".join([[abstract] + sections]),
|
| 142 |
_SUMMARY: summary,
|
| 143 |
+
"section_headings": "\n".join(["Abstract"] + d["headings"]),
|
| 144 |
"keywords": "\n".join(d["keywords"]),
|
| 145 |
"year": d["year"],
|
| 146 |
"title": d["title"]
|