Details¶

This page describe how the dataset is derived and the description of it.

Data Sources¶

The citations are obtained from the following sources:

CrossRef via DOI obtained from Open Academic Graph
JSTOR Sample Dataset (not accessible anymore)
PubMed (2019 Baseline)

CSL Styles¶

In total, 17 styles have been employed. The table below summarises the number of reference strings available in the dataset for each style.

Number of Strings By CSL Style¶
CSL Style	Number of Reference Strings
Annual Reviews	placeholder
APA 6th edition	placeholder
Cambridge University Press	placeholder
Chicago	placeholder
Current Opinion	placeholder
Elsevier (Harvard)	placeholder
Elsevier (Vancouver)	placeholder
IEEE	placeholder
MLA 7th edition	placeholder
Nature	placeholder
University of New South Wales (Oxford)	placeholder
Springer Humanities	placeholder
Springer MathPhys	placeholder
Springer (Vancouver)	placeholder
Taylor and Francis (Harvard)	placeholder
Wiley-VCH Books	placeholder

BibTeX Entry Types¶

The table below summarises the number of reference strings available for each BibTeX entry type.

Number of Strings By BibTeX Entry Types¶
Entry Type	Number of Reference Strings
article	placeholder
book	placeholder
inbook	placeholder
incollection	placeholder
inproceedings	placeholder
misc	placeholder
phdthesis	placeholder
techreport	placeholder

Data Format¶

The data are stored as JSON lines in each file. Each line of the files represents a citation rendered in a specified CSL style with its corresponding annotated sequence.

An example (annotated by segment) with its metadata such as the source, document type and the style the citation is rendered¶

{
   "style": "apa",
   "doc_type": "article",
   "source": "crossref",
   "data": "<author>Watson, J. D., & Crick, F. H. C.</author> <year>(1953).</year> <title>Molecular structure of nucleic acids: A structure for deoxyribose nucleic acid.</title> <container-title>Nature,</container-title> <volume>171</volume> <issue>(4356),</issue> <page>737\\u2013738.</page> <DOI>https://doi.org/10.1038/171737a0</DOI>"
}

Important

Not all tokens are enclosed within the tags. These should be labelled as O (according to tagging scheme).