README.md

metadata

license: cc-by-sa-3.0
config_names:
  - Abbreviation equality
  - Adjective inflection analogy
  - Clinical analogy
  - Clinical similarity
  - Noun inflection analogy
  - UMNSRS relatedness
  - UMNSRS similarity
  - Verb inflection analogy
configs:
  - config_name: Abbreviation equality
    data_files:
      - split: train
        path: Abbreviation equality/train*
  - config_name: Adjective inflection analogy
    data_files:
      - split: train
        path: Adjective inflection analogy/train*
  - config_name: Clinical analogy
    data_files:
      - split: train
        path: Clinical analogy/train*
  - config_name: Clinical similarity
    data_files:
      - split: train
        path: Clinical similarity/train*
  - config_name: Noun inflection analogy
    data_files:
      - split: train
        path: Noun inflection analogy/train*
  - config_name: UMNSRS relatedness
    data_files:
      - split: train
        path: UMNSRS relatedness/train*
  - config_name: UMNSRS similarity
    data_files:
      - split: train
        path: UMNSRS similarity/train*
  - config_name: Verb inflection analogy
    data_files:
      - split: train
        path: Verb inflection analogy/train*

Danish medical word embedding evaluation

The development of the dataset is described further in our paper.

Citing

@inproceedings{laursen-etal-2023-benchmark,
    title = "Benchmark for Evaluation of {D}anish Clinical Word Embeddings",
    author = "Laursen, Martin Sundahl  and
      Pedersen, Jannik Skyttegaard  and
      Vinholt, Pernille Just  and
      Hansen, Rasmus S{\o}gaard  and
      Savarimuthu, Thiusius Rajeeth",
    editor = "Derczynski, Leon",
    booktitle = "Northern European Journal of Language Technology, Volume 9",
    year = "2023",
    address = {Link{\"o}ping, Sweden},
    publisher = {Link{\"o}ping University Electronic Press},
    url = "https://aclanthology.org/2023.nejlt-1.4",
    doi = "https://doi.org/10.3384/nejlt.2000-1533.2023.4132",
    abstract = "In natural language processing, benchmarks are used to track progress and identify useful models. Currently, no benchmark for Danish clinical word embeddings exists. This paper describes the development of a Danish benchmark for clinical word embeddings. The clinical benchmark consists of ten datasets: eight intrinsic and two extrinsic. Moreover, we evaluate word embeddings trained on text from the clinical domain, general practitioner domain and general domain on the established benchmark. All the intrinsic tasks of the benchmark are publicly available.",
}