Skip to content

Types of Metadata

Items in the Anthology can have various kinds of metadata, which are listed and described in detail in the API documentation for their respective object; e.g. papers, volumes, events, SIGs, venues, etc.

Plain metadata fields

Some metadata fields are simply strings or lists of strings; for example:

paper = anthology.get("2022.acl-long.220")
paper.bibkey           # is 'kitaev-etal-2022-learned'
paper.doi              # is '10.18653/v1/2022.acl-long.220'
paper.year             # is '2022'
paper.awards           # is ['Best Paper']

However, many of them are optional, and can be None or empty.

paper.attachments      # is {}
paper.language         # is None

Markup text

Two of the most common field types, titles and abstracts, can contain markup text. They support a limited set of markup tags to denote, e.g., bold and italics, TeX math expressions, URLs, protected case (i.e. marking capitals that should never be lowercased in bibliographic references), and potentially more in the future.

This means that such metadata fields don't contain strings, but rather MarkupText instances:

>>> paper = anthology.get("2022.acl-short.58")
>>> paper.title
MarkupText('S<span class="tex-math"><sup>4</sup></span>-Tuning: A Simple Cross-lingual Sub-network Tuning Method')

To work with these as Unicode strings, you need to explicitly convert them: 

>>> str(paper.title)
'S4-Tuning: A Simple Cross-lingual Sub-network Tuning Method'

Alternatively, they can be converted into HTML or LaTeX representations:

>>> paper.title.as_html()
'S<span class="tex-math"><sup>4</sup></span>-Tuning: A Simple Cross-lingual Sub-network Tuning Method'
>>> paper.title.as_latex()
'S$^4$-Tuning: A Simple Cross-lingual Sub-network Tuning Method'

File references

Some metadata fields refer to files, the most common of which is the PDF file for a paper:

>>> paper = anthology.get("2021.eacl-main.89")
>>> paper.pdf
PDFReference(name='2021.eacl-main.89', checksum='27663bea')

These are always classes that inherit from FileReference. As you can see above, they usually1 give only the base filename. You can, however, easily obtain the full URL for these:

>>> paper.pdf.url
'https://aclanthology.org/2021.eacl-main.89.pdf'

Some other places where you might encounter file references are attachments or event links:

>>> paper.attachments
{
  'Software': AttachmentReference(name='2021.eacl-main.89.Software.zip', checksum='ca8afe3f'),
  'Dataset': AttachmentReference(name='2021.eacl-main.89.Dataset.txt', checksum='07778fe5')
}
>>> anthology.get_event("acl-2022").links
{
  'website': AttachmentReference(name='https://2022.aclweb.org', checksum=None),
  'handbook': AttachmentReference(name='2022.acl.handbook.pdf', checksum=None)
}

Most file references also include the checksum field, which is a CRC32 checksum of the respective file that can be used to verify if it was downloaded correctly.

  1. For all materials hosted on the ACL Anthology web server. ↩