SoFairOA/somesci_dataset

SoMeSci Dataset

SoMeSci dataset is a dataset of software mentions in scientific papers. This dataset was converted from the original GitHUB repository SoMeSci.

The original upload is containing some documents without content. These documents are removed from this dataset.

Configurations

This dataset is available in three configurations that corresponds to original SoMeSci subsets:

• pubmed_fulltext (default)
• plos_methods
• plos_sentences

Fields description

All configurations contain the following fields:

• id: The unique identifier of the example. Corresponds to sample file name in the original dataset.

• text: The text of the paragraph or the full text of the paper.

• tokens: Tokenized text for which the labels are provided.

• start_offsets: The start character offsets of the tokens in the text.

• labels: The labels of the tokens. The labels are in the BIO format:

  • 0: O
  • 1: B-PROGRAMMINGENVIRONMENT
  • 2: I-PROGRAMMINGENVIRONMENT
  • 3: B-SOFTWARECOREFERENCE
  • 4: I-SOFTWARECOREFERENCE
  • 5: B-APPLICATION
  • 6: I-APPLICATION
  • 7: B-PLUGIN
  • 8: I-PLUGIN
  • 9: B-OPERATINGSYSTEM
  • 10: I-OPERATINGSYSTEM
  • 11: B-ALTERNATIVENAME
  • 12: I-ALTERNATIVENAME
  • 13: B-URL
  • 14: I-URL
  • 15: B-ABBREVIATION
  • 16: I-ABBREVIATION
  • 17: B-LICENSE
  • 18: I-LICENSE
  • 19: B-CITATION
  • 20: I-CITATION
  • 21: B-VERSION
  • 22: I-VERSION
  • 23: B-EXTENSION
  • 24: I-EXTENSION
  • 25: B-RELEASE
  • 26: I-RELEASE
  • 27: B-DEVELOPER
  • 28: I-DEVELOPER

• software: Document level software mentions. Contains the following fields:

  • name
  • type
  • alternativename
  • url
  • abbreviation
  • license
  • citation
  • version
  • extension
  • release
  • developer

  Disambiguation is done on lexical level. Meaning that two software names are considered different if they have different normalized forms.