|
--- |
|
license: apache-2.0 |
|
--- |
|
|
|
|
|
# CSAbstruct |
|
|
|
CSAbstruct was created as part of ["Pretrained Language Models for Sequential Sentence Classification"][1]. |
|
|
|
It contains 2,189 manually annotated computer science abstracts with sentences annotated according to their rhetorical roles in the abstract, similar to the [PUBMED-RCT][2] categories. |
|
|
|
|
|
## Dataset Construction Details |
|
|
|
CSAbstruct is a new dataset of annotated computer science abstracts with sentence labels according to their rhetorical roles. |
|
The key difference between this dataset and [PUBMED-RCT][2] is that PubMed abstracts are written according to a predefined structure, whereas computer science papers are free-form. |
|
Therefore, there is more variety in writing styles in CSAbstruct. |
|
CSAbstruct is collected from the Semantic Scholar corpus [(Ammar et al., 2018)][3]. |
|
Each sentence is annotated by 5 workers on the [Figure-eight platform][4], with one of 5 categories `{BACKGROUND, OBJECTIVE, METHOD, RESULT, OTHER}`. |
|
|
|
We use 8 abstracts (with 51 sentences) as test questions to train crowdworkers. |
|
Annotators whose accuracy is less than 75% are disqualified from doing the actual annotation job. |
|
The annotations are aggregated using the agreement on a single sentence weighted by the accuracy of the annotator on the initial test questions. |
|
A confidence score is associated with each instance based on the annotator initial accuracy and agreement of all annotators on that instance. |
|
We then split the dataset 75%/15%/10% into train/dev/test partitions, such that the test set has the highest confidence scores. |
|
Agreement rate on a random subset of 200 sentences is 75%, which is quite high given the difficulty of the task. |
|
Compared with [PUBMED-RCT][2], our dataset exhibits a wider variety of writ- ing styles, since its abstracts are not written with an explicit structural template. |
|
|
|
## Dataset Statistics |
|
|
|
| Statistic | Avg ± std | |
|
|--------------------------|-------------| |
|
| Doc length in sentences | 6.7 ± 1.99 | |
|
| Sentence length in words | 21.8 ± 10.0 | |
|
|
|
| Label | % in Dataset | |
|
|---------------|--------------| |
|
| `BACKGROUND` | 33% | |
|
| `METHOD` | 32% | |
|
| `RESULT` | 21% | |
|
| `OBJECTIVE` | 12% | |
|
| `OTHER` | 03% | |
|
|
|
## Citation |
|
|
|
If you use this dataset, please cite the following paper: |
|
|
|
``` |
|
@inproceedings{Cohan2019EMNLP, |
|
title={Pretrained Language Models for Sequential Sentence Classification}, |
|
author={Arman Cohan, Iz Beltagy, Daniel King, Bhavana Dalvi, Dan Weld}, |
|
year={2019}, |
|
booktitle={EMNLP}, |
|
} |
|
``` |
|
|
|
[1]: https://aclanthology.org/D19-1383 |
|
[2]: https://arxiv.org/abs/1710.06071 |
|
[3]: https://aclanthology.org/N18-3011/ |
|
[4]: https://www.figure-eight.com/ |
|
|
