metadata
license: cc-by-4.0
Annotation data columns
ea_justification_annotations.csv contains 1,176 rows and 83 columns.
| Column group | Count | Labels/range | Description |
|---|---|---|---|
dataset, item_id, sample_rank, annotation_batch |
4 | dataset: hansard or qam; rank: positive integer |
Dataset source, item identifier, sampling order, and annotation batch. |
topic, argument, original_argument, argument_group |
4 | argument_group: real or synthetic |
Topic and argument text, original argument, and generation status. |
emotional_appeal_injector_model, source_emotional_appeal_type, source_emotional_appeal_strength |
3 | Model/human provenance; seven EA types; strength 2–5 |
Argument-generation provenance and intended appeal label. |
context, original_context, context_group |
3 | context_group: real or synthetic |
Presented context, original context, and perturbation status. |
context_perturber_model, context_aspect, context_level |
3 | Aspect: mag, prob, ris, sprox, tprox; level: LOW, MEDIUM, HIGH |
Perturbation model, aspect, and target level. |
context_pair_id, context_pair_role |
2 | Role: original or perturbed |
Original/perturbed context pairing. |
annotator_{1..5}_id |
5 | Anonymized categorical IDs | Annotator identifiers. |
annotator_{1..5}_context_support |
5 | Integer −3–3 |
Argument-context support rating. |
annotator_{1..5}_emotional_appeal_type |
5 | Nine EA-type labels | Primary emotional appeal. |
annotator_{1..5}_emotional_appeal_strength |
5 | Integer 1–5 |
Emotional-appeal strength. |
annotator_{1..5}_emotion_type_fit |
5 | Integer 1–5 or not_applicable |
Fit between the argument and identified emotion type. |
annotator_{1..5}_evidential_proportionality |
5 | Integer 1–5 or not_applicable |
Proportionality of the appeal to its evidence/context. |
annotator_{1..5}_evidential_proportionality_reasons |
5 | Pipe-separated selection from nine reason labels | Reasons for the evidential-proportionality rating. |
annotator_{1..5}_normative_acceptability |
5 | Integer 1–5 or not_applicable |
Normative acceptability rating. |
annotator_{1..5}_overall_justification |
5 | Integer 1–5 or not_applicable |
Overall justification rating. |
paper_top3_annotator_{1..3}_id |
3 | Anonymized categorical IDs | Annotators selected per batch for the paper analysis. |
{dimension}_all5_majority |
8 | Same domain as the corresponding dimension | Majority aggregates over all available annotators. |
{dimension}_paper_top3_majority |
8 | Same domain as the corresponding dimension | Majority aggregates over the selected top three. |