orkg-R0: A Dataset of Structured Summaries for the R0 estimate of Infectious Diseases from Complex Scientific Abstracts
收藏NIAID Data Ecosystem2026-05-01 收录
下载链接：
https://zenodo.org/record/8068441
下载链接
链接失效反馈官方服务：
资源简介：
This dataset is a curated dataset obtained by filtering, cleaning and manually annotating the metadata file available at CORD_19 dataset (https://allenai.org/data/cord-19). It contains structured summaries for the R0 estimate of infectious diseases from scientific abstracts.
The main data directory contains two subdirectories, "raw" sub-folder where it holds the train, test, dev splits of the annotated data.
The "processed" subdirectory contains train, test, and dev JSON files filled in a sub-selection of "Templates for FLAN." prompts.
The sub-selection is :
Drop and Squad_v2 templates. template number 8 from Drop and template number 3 from Squad_v2 have been excluded among all splits.
Templates 9 and 10 from Drop have been just used in the training sets.
Two main dataset types are included in this repository: Text_based and Json_based.
The "dev_templated_files" subdirectory contains two subdirectories of "text" and "json".
The "text" sub-folder contains the raw "dev" split filled in all suitable templates for dev where the responses are in the defined structured text_based format.
The "json" sub-folder contains the raw "dev" split filled in all suitable templates for dev where the responses are in the defined structured json_based format.
The "test_templated_files" subdirectory contains two subdirectories of "text" and "json".
The "text" sub-folder contains the raw "test" split filled in all suitable templates for dev where the responses are in the defined structured text_based format.
The "json" sub-folder contains the raw "test" split filled in all suitable templates for dev where the responses are in the defined structured json_based format.
The "train_templated_files" Subdirectory contains subdirectories each representing a train dataset obtained using the specific templates.
it contains 20 different train sets each having 2 json_based and text_based versions, resulting in 40 different training sets.
创建时间：
2023-06-25
