ArgPremises — Corpus of Matched Claims with Implicit Premises

Version: 1.0
Release date: June 28, 2016

1 Description

ArgPremises is a dataset of implicit premises manually annotated on matching claims. The dataset is created to explore what are the implied premises users make when expressing support for a claim. Additionally, it was used to help solve the claim matching task; using implcit premise information whilst identifying whether a pair of claims expresses the same argument. The task and the dataset are described in:

If you use the ComArg dataset for your own work, please cite the above paper. The BibTeX citation is:

@InProceedings{boltuzic2014back,
  author    = {Boltu\v{z}i\'{c}, Filip  and  \v{S}najder, Jan},
  title     = {Fill the Gap! Analyzing Implicit Premises between Claims from Online Debates},
  booktitle = {Proceedings of the 3rd Workshop on Argumentation Mining},
  month     = {August},
  year      = {2016},
  address   = {Berlin},
  publisher = {Association for Computational Linguistics}
}

2 Dataset

The dataset is available from here: TakeLab-argpremises.tar.gz.
The archive contains 494 claim pairs and 3977 respective premises on topics:

Marijuana Legalization (MA)
Gay Rights (GA)
Abortion Legalization (AB)
Obama Presidency (OB)

The schema of the claim pairs and premises is as follows:

{ "claim_pairs": [
        {
            "topic": "..", 
            "post_id": "..", 
            "premises": [
                {
                    "anotator": "..", 
                    "premise_id": "..", 
                    "premise_text": ".."
                },
                ... 
                {
                    "anotator": "..", 
                    "premise_id": "..", 
                    "premise_text": ".."
                }
            ], 
            "user_claim_text": "..", 
            "main_claim_text": "..", 
            "main_claim_stance": "..", 
            "main_claim_id": "..", 
            "claim_pair_id": "..", 
            "user_claim_id": ".."
        }, 
}

3 License

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.