license: cc
Dataset Card for Dataset Name
Small genetics-related text dataset based on 23200 ArXiv abstact records and 111 Wikipedia pages.
Dataset Details
Dataset Description
Dataset was produced using the python scripts you will find in this GitHub repository.
It represents a collection of genetics-related text data taken from ArXiv abstracts dataset and Wikipedia.
Dataset holds a total of 23311 text records, 23200 of which belonging to categories q-bio.BM, q-bio.GN, q-bio.MN, q-bio.PE and q-bio.SC from ArXiv dataset and the remainder 111 represent text from pages in Genetics category on English Wikipeda
Dataset is in jsonl format and its size is 28 MB.
- Curated by: Astra Clelia Bertelli
- Language(s) (NLP): English
- License: Creative Commons family
Dataset Source
Data and scripts are placed in a wider-scope repository on GitHub
- Repository: AstraBert/model_training
Uses
Dataset can be use for finetuning text-generation models on Genetics and Genetics-related contents.
It can also be used for research and teaching purposes.
Direct Use
Find a direct application of the dataset in this model
Out-of-Scope Use
Dataset should not be used to create or sell misleading content, spread disinformation and break the laws.
Responsibility of out-of-scope uses is completely up to the user.
Dataset Structure
Dataset is composed by a jsonl file with 23331 lines, each described by one field named "text".
Dataset Creation
Curation Rationale
Dataset was created to finetune language/text-generation models on Genetics.
Source Data
As stated before, source data are:
Data Collection and Processing
- Data from Wikipedia were collected via python interaction with Wikipedia API; find the code here
- Data from ArXiv dataset were first downloaded from the dedicated Kaggle page and then processed with the following code
Personal and Sensitive Information
There are no personal or sensitive information. All data are public and freely accessible.
Bias, Risks, and Limitations
Being a dataset built on pre-print paper abstracts and on Wikipedia data, review process of them may have failed to detect errors or ambiguous information.
There is no evident or intrinsic bias, except for the dataset being mostly research-oriented (most of the data comes from research papers).
Apart from possibly containing errors (and thus outputting wrong information if fed to a textgen/language model), there should not be any intrinsic risks connected to the dataset.
Recommendations
Users should comply with dataset suggested uses and should not employ the dataset for out-of-scope uses.
Users should be aware of the risks, biases and limitations of the dataset as stated before.