Datasets:
Dataset Overview
Where to find the data and its documentation
What is the webpage for the dataset (if it exists)?
https://github.com/AshOlogn/Paragraph-level-Simplification-of-Medical-Texts
What is the link to where the original dataset is hosted?
https://github.com/AshOlogn/Paragraph-level-Simplification-of-Medical-Texts
What is the link to the paper describing the dataset (open access preferred)?
https://aclanthology.org/2021.naacl-main.395/
Provide the BibTex-formatted reference for the dataset. Please use the correct published version (ACL anthology, etc.) instead of google scholar created Bibtex.
@inproceedings{devaraj-etal-2021-paragraph,
title = "Paragraph-level Simplification of Medical Texts",
author = "Devaraj, Ashwin and
Marshall, Iain and
Wallace, Byron and
Li, Junyi Jessy",
booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
month = jun,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.naacl-main.395",
doi = "10.18653/v1/2021.naacl-main.395",
pages = "4972--4984",
}
If known, provide the name of at least one person the reader can contact for questions about the dataset.
Ashwin Devaraj
If known, provide the email of at least one person the reader can contact for questions about the dataset.
Does the dataset have an active leaderboard?
no
Languages and Intended Use
Is the dataset multilingual?
no
What languages/dialects are covered in the dataset?
English
What is the license of the dataset?
cc-by-4.0: Creative Commons Attribution 4.0 International
What is the intended use of the dataset?
The intended use of this dataset is to train models that simplify medical text at the paragraph level so that it may be more accessible to the lay reader.
What primary task does the dataset support?
Simplification
Provide a short description of the communicative goal of a model trained for this task on this dataset.
A model trained on this dataset can be used to simplify medical texts to make them more accessible to readers without medical expertise.
Credit
In what kind of organization did the dataset curation happen?
academic
Name the organization(s).
The University of Texas at Austin, King's College London, Northeastern University
Who created the original dataset? List the people involved in collecting the dataset and their affiliation(s).
Ashwin Devaraj (The University of Texas at Austin), Iain J. Marshall (King's College London), Byron C. Wallace (Northeastern University), Junyi Jessy Li (The University of Texas at Austin)
Who funded the data creation?
National Institutes of Health (NIH) grant R01-LM012086, National Science Foundation (NSF) grant IIS-1850153, Texas Advanced Computing Center (TACC) computational resources
Who contributed to the data card and adding the dataset to GEM? List the people+affiliations involved in creating this data card and who helped integrate this dataset into GEM.
Ashwin Devaraj (The University of Texas at Austin)
Structure
List and describe the fields present in the dataset.
gem_id: string, a unique identifier for the example doi: string, DOI identifier for the Cochrane review from which the example was generated source: string, an excerpt from an abstract of a Cochrane review target: string, an excerpt from the plain-language summary of a Cochrane review that roughly aligns with the source text
Provide a JSON formatted example of a typical instance in the dataset.
{
"gem_id": "gem-cochrane-simplification-train-766",
"doi": "10.1002/14651858.CD002173.pub2",
"source": "Of 3500 titles retrieved from the literature, 24 papers reporting on 23 studies could be included in the review. The studies were published between 1970 and 1997 and together included 1026 participants. Most were cross-over studies. Few studies provided sufficient information to judge the concealment of allocation. Four studies provided results for the percentage of symptom-free days. Pooling the results did not reveal a statistically significant difference between sodium cromoglycate and placebo. For the other pooled outcomes, most of the symptom-related outcomes and bronchodilator use showed statistically significant results, but treatment effects were small. Considering the confidence intervals of the outcome measures, a clinically relevant effect of sodium cromoglycate cannot be excluded. The funnel plot showed an under-representation of small studies with negative results, suggesting publication bias. There is insufficient evidence to be sure about the efficacy of sodium cromoglycate over placebo. Publication bias is likely to have overestimated the beneficial effects of sodium cromoglycate as maintenance therapy in childhood asthma.",
"target": "In this review we aimed to determine whether there is evidence for the effectiveness of inhaled sodium cromoglycate as maintenance treatment in children with chronic asthma. Most of the studies were carried out in small groups of patients. Furthermore, we suspect that not all studies undertaken have been published. The results show that there is insufficient evidence to be sure about the beneficial effect of sodium cromoglycate compared to placebo. However, for several outcome measures the results favoured sodium cromoglycate."
}
Describe and name the splits in the dataset if there are more than one.
train: 3568 examples validation: 411 examples test: 480 examples
Dataset in GEM
Rationale
What does this dataset contribute toward better generation evaluation and why is it part of GEM?
This dataset is the first paragraph-level simplification dataset published (as prior work had primarily focused on simplifying individual sentences). Furthermore, this dataset is in the medical domain, which is an especially useful domain for text simplification.
Do other datasets for the high level task exist?
no
Does this dataset cover other languages than other datasets for the same task?
no
What aspect of model ability can be measured with this dataset?
This dataset measures the ability for a model to simplify paragraphs of medical text through the omission non-salient information and simplification of medical jargon.
GEM Additional Curation
Has the GEM version of the dataset been modified in any way (data, processing, splits) from the original curated data?
no
Does GEM provide additional splits to the dataset?
no
Getting Started
Previous Results
Previous Results
What aspect of model ability can be measured with this dataset?
This dataset measures the ability for a model to simplify paragraphs of medical text through the omission non-salient information and simplification of medical jargon.
What metrics are typically used for this task?
Other: Other Metrics
Definitions of other metrics
SARI BLEU
List and describe the purpose of the metrics and evaluation methodology (including human evaluation) that the dataset creators used when introducing this task.
SARI - measures quality of text simplification BLEU - precision-based method to used to score quality of machine translation
Are previous results available?
yes
What are the most relevant previous results for this task/dataset?
The paper which introduced this dataset trained BART models (pretrained on XSum) with unlikelihood training to produce simplification models achieving maximum SARI and BLEU scores of 40 and 43 respectively.
Dataset Curation
Original Curation
Is the dataset aggregated from different data sources?
no
Language Data
Was the text validated by a different worker or a data curator?
not validated
Were text instances selected or filtered?
not filtered
Structured Annotations
Does the dataset have additional annotations for each instance?
none
Was an annotation service used?
no
Consent
Was there a consent policy involved when gathering the data?
no
Private Identifying Information (PII)
Does the source language data likely contain Personal Identifying Information about the data creators or subjects?
yes/very likely
Did the curators use any automatic/manual method to identify PII in the dataset?
no identification
Maintenance
Does the original dataset have a maintenance plan?
no
Broader Social Context
Previous Work on the Social Impact of the Dataset
Are you aware of cases where models trained on the task featured in this dataset ore related tasks have been used in automated systems?
no
Impact on Under-Served Communities
Does this dataset address the needs of communities that are traditionally underserved in language technology, and particularly language generation technology? Communities may be underserved for exemple because their language, language variety, or social or geographical context is underepresented in NLP and NLG resources (datasets and models).
yes
Describe how this dataset addresses the needs of underserved communities.
This dataset can be used to simplify medical texts that may otherwise be inaccessible to those without medical training.
Discussion of Biases
Are there documented social biases in the dataset? Biases in this context are variations in the ways members of different social categories are represented that can have harmful downstream consequences for members of the more disadvantaged group.
unsure
Does the distribution of language producers in the dataset accurately represent the full distribution of speakers of the language world-wide? If not, how does it differ?
The dataset was generated from abstracts and plain-language summaries of medical literature reviews that were written by medical professionals and thus does was not generated by people representative of the entire English-speaking population.
Considerations for Using the Data
PII Risks and Liability
Licenses
Known Technical Limitations
Describe any known technical limitations, such as spurrious correlations, train/test overlap, annotation biases, or mis-annotations, and cite the works that first identified these limitations when possible.
The main limitation of this dataset is that the information alignment between the abstract and plain-language summary is often rough, so the plain-language summary may contain information that isn't found in the abstract. Furthermore, the plain-language targets often contain formulaic statements like "this evidence is current to [month][year]" not found in the abstracts. Another limitation is that some plain-language summaries do not simplify the technical abstracts very much and still contain medical jargon.
When using a model trained on this dataset in a setting where users or the public may interact with its predictions, what are some pitfalls to look out for? In particular, describe some applications of the general task featured in this dataset that its curation or properties make it less suitable for.
The main pitfall to look out for is errors in factuality. Simplification work so far has not placed a strong emphasis on the logical fidelity of model generations with the input text, and the paper introducing this dataset does not explore modeling techniques to combat this. These kinds of errors are especially pernicious in the medical domain, and the models introduced in the paper do occasionally alter entities like disease and medication names.