|
--- |
|
license: cc-by-4.0 |
|
language: |
|
- en |
|
tags: |
|
- biology |
|
- plant science |
|
- named entity recognition |
|
- relation extraction |
|
pretty_name: PICKLE |
|
--- |
|
# PICKLE |
|
PICKLE is the dataset associated with the manuscript *In a PICKLE: A gold standard entity and relation corpus for the molecular plant sciences*. The code repository associated with this dataset can be found [here](https://github.com/serenalotreck/pickle-corpus-code). |
|
|
|
## Format specification |
|
This dataset is formatted according to the [specifications for use with the DyGIE++ architecture](https://github.com/dwadden/dygiepp/blob/master/doc/data.md). |
|
**NOTE:** At this time, the dataset will throw a `JSONDecodeError` when used with `load_datasets` (see [#6460 on `datasets`](https://github.com/huggingface/datasets/issues/6460)). In the meantime, you can access the data by downloading the `.jsonl` files directly from the GUI, and importing to Python with the following code: |
|
``` |
|
import jsonlines |
|
with jsonlines.open('train.jsonl') as reader: |
|
train = [] |
|
for obj in reader: |
|
train.append(obj) |
|
``` |
|
|
|
## Dataset details |
|
There are a total of 250 documents in `all.jsonl`, split up into 68%/12%/20% train/dev/test. Each document is an abstract from a scientific paper in the search results for the terms "gibberellic acid" and "jasmonic acid". There are 6,245 entity and 2,149 relation annotations across the 250 documents. |
