deepa2 Datasets Collection

Dataset Summary

This is a growing, curated collection of deepa2 datasets, i.e. datasets that contain comprehensive logical analyses of argumentative texts. The collection comprises:

• datasets that are built from existing NLP datasets by means of the deepa2 bake tool.
• original deepa2 datasets specifically created for this collection.

The tool deepa2 serve may be used to render the data in this collection as text2text examples.

Supported Tasks and Leaderboards

For each of the tasks tagged for this dataset, give a brief description of the tag, metrics, and suggested models (with a link to their HuggingFace implementation if available). Give a similar description of tasks that were not covered by the structured tag set (repace the task-category-tag with an appropriate other:other-task-name ).

• conditional-text-generation : The dataset can be used to train models to generate a fully reconstruction of an argument from a source text, making, e.g., its implicit assumptions explicit.
• structure-prediction : The dataset can be used to train models to formalize sentences.
• text-retrieval : The dataset can be used to train models to extract reason statements and conjectures from a given source text.

Languages

English. Will be extended to cover other languages in the futures.

Dataset Structure

Sub-Datasets

This collection contains the following deepa2 datasets:

• esnli : created from e-SNLI with deepa2 bake as described here .
• enbank ( task_1 , task_2 ): created from Entailment Bank with deepa2 bake as described here .
• argq : created from IBM-ArgQ with deepa2 bake as described here .
• argkp : created from IBM-KPA with deepa2 bake as described here .
• aifdb ( moral-maze , us2016 , vacc-itc ): created from AIFdb with deepa2 bake as described here .
• aaac ( aaac01 and aaac02 ): original, machine-generated contribution; based on an an improved and extended algorithm that backs https://huggingface.co/datasets/debatelab/aaac .

Data Instances

see: https://github.com/debatelab/deepa2/tree/main/docs

Data Fields

see: https://github.com/debatelab/deepa2/tree/main/docs

Data Splits

Each sub-dataset contains three splits: train , validation , and test .

Dataset Creation

Curation Rationale

Many NLP datasets focus on tasks that are relevant for logical analysis and argument reconstruction. This collection is the attempt to unify these resources in a common framework.

Source Data

See: Sub-Datasets

Additional Information

Dataset Curators

Gregor Betz, KIT; Kyle Richardson, Allen AI

Licensing Information

We re-distribute the the imported sub-datasets under their original license:

Citation Information

@article{betz2021deepa2,
      title={DeepA2: A Modular Framework for Deep Argument Analysis with Pretrained Neural Text2Text Language Models}, 
      author={Gregor Betz and Kyle Richardson},
      year={2021},
      eprint={2110.01509},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}