List of Tasks

The list below links to data statements [1, 2] for each of the datasets that are part of GEM tasks. The template used to produce the statements and a guide on how to write them can be found here: [download template] [view guide].

  • MLSumsummarization
    Large-scale multilingual dataset for evaluating abstractive summarization
  • XSumSummarization
    Large scale monolingual dataset for evaluating extreme summarization.
  • WikiLinguaSummarization
    Large-scale multilingual dataset for evaluating cross-lingual abstractive summarization
  • WebNLGStructure-to-text
    The WebNLG dataset is a large bi-lingual dataset with crowdsourced reference texts and a rather large variety of knowledge in the inputs. A web-based evaluation platform is already existing.
  • CommonGenStructure-to-text
    A medium sized corpus with a unique reasoning challenge and interesting evaluation possibilities.
  • E2EStructure-to-Text
    One of the largest limited-domain NLG datasets and is frequently used as a data-to-text generation benchmark.
  • DARTStructure-to-Text
    Hierarchical, structured format with its open-domain nature
  • Czech RestaurantStructure-to-Text
    One of a few non-English data-to-text datasets in a well-known domain, covering a morphologically rich language.
  • ToTToStructure-To-Text
    Controlled Table2Text task with non-divergent, annotator-revised text outputs
  • Wiki-AutoSimplification
    Wiki-Auto is the largest open text simplification dataset currently available. For GEM, Wiki-Auto acts as the training set.
  • TURKCorpusSimplification
    TURKCorpus is a high-quality simplification dataset where each source sentence is associated with 8 human-written simplifications.
  • ASSETSimplification
    ASSET is a high quality simplification dataset where each source (not simple) sentence is associated with 10 human-written simplifications.
  • Schema-Guided DialogDialog
    Modeling task-oriented dialog.