---
language:
- it
license: other
license_name: copyrighted-research-purposes
license_link: LICENSE
tags:
- word-game
- rebus
- italian
- word-puzzle
- crossword
language_creators:
- expert-generated
pretty_name: eureka-rebus
size_categories:
- 100K<n<1M
source_datasets:
- Kamyar-zeinalipour/ITA_CW
task_categories:
- text2text-generation
configs:
- config_name: verbalized_rebus
  data_files:
  - split: train
    path: verbalized_rebus.csv
- config_name: eureka_rebus
  data_files:
  - split: train
    path: rebus.csv
  default: true
- config_name: llm_sft
  data_files:
  - split: train
    path: train.jsonl
  - split: test
    path: 
    - id_test.jsonl
    - ood_test.jsonl
  - split: id_test
    path: id_test.jsonl
  - split: ood_test
    path: ood_test.jsonl
extra_gated_prompt: "### Verbalized Rebus Dataset License Agreement\
    \ \n\n\"Agreement\" means the terms and conditions for\
    \ use, reproduction, distribution and modification of the materials set forth\
    \ herein.\n\n1. Data Source and Redistribution\n\nThe contents of the 'rebus.csv'\
    \ file are extracted directly from the open database 'Eureka5', hosted by the\
    \ Associazione Culturale \"Biblioteca Enigmistica Italiana - G. Panini\" at the\
    \ address http://www.eureka5.it. This data is redistributed with the explicit\
    \ approval of the association for research purposes only.\n\n2. Copyright of\
    \ Individual Entries\n\nThe rights to each entry in this collection\
    \ are the property of the respective copyright holders. These copyright holders\
    \ are identified by the acronym in the 'RIVISTA' field in the 'rebus.csv' file.\
    \ A comprehensive list of these acronyms and their corresponding publications\
    \ is available at http://www.eureka5.it/public/Leg_RivisteSearchCrud.\n\n3.\
    \ Derivative Works\n\nOther files in this repository contain derivatives of\
    \ the original data, specifically derivatives of the 'PRIMALET' field from 'rebus.csv'\
    \ where words have been replaced by crossword definitions. All such derivative\
    \ works remain under copyright protection. These derivative works, like the\
    \ original data, are to be used exclusively for non-commercial research purposes\
    \ in accordance with Article 70 of the Italian Copyright Law. Any further derivative\
    \ work building upon these resources should remain licensed with a non-commercial\
    \ license allowing research applications.\n\n4. Terms of Use\n\na. Non-commercial\
    \ Use Only: All data and derivative works in this repository may be used solely\
    \ for non-commercial research purposes.\nb. Attribution: Users must provide appropriate\
    \ attribution to the original copyright holders and the Associazione Culturale\
    \ \"Biblioteca Enigmistica Italiana - G. Panini\" when using these data for research\
    \ purposes.\ne. Derivative Works: The creation of derivative works is permitted\
    \ only for research purposes and such works must be shared under terms no less\
    \ restrictive than this license.\n\n5. Disclaimer\n\nThis data is provided \"as\
    \ is\" without any warranties, express or implied. The licensors make no representations\
    \ or warranties regarding the accuracy, completeness, or fitness for a particular\
    \ purpose of the data.\n\n6. Right of Removal\n\nWe reserve the right to remove\
    \ any part of the data from the repository at any time, particularly in the event\
    \ that a copyright holder requests its removal.\n\n7. Termination\n\nThis license\
    \ will terminate automatically if you fail to comply with the terms stated herein.\
    \ Upon termination, you must destroy all copies of the data in your possession.\n\n8.\
    \ Governing Law\n\nThis license shall be governed by and interpreted in accordance\
    \ with the laws of Italy, particularly in relation to copyright law and fair\
    \ use for research purposes.\n\nBy accessing or using the data in this repository,\
    \ you agree to be bound by the terms of this license. If you do not agree to these\
    \ terms, you are not permitted to access or use the data.\n\nFor any questions\
    \ regarding this license or requests for additional permissions, please contact\
    \ gabriele.sarti996@gmail.com.\n\nLast updated: July 30th, 2024"
extra_gated_fields:
  I understand that the rights over EurekaRebus contents belong to the respective copyright holders: checkbox
  I agree to use this dataset exclusively for non-commercial research purposes: checkbox
  I agree to comply with the data license for this dataset: checkbox
---

# Dataset Card for EurekaRebus

## Dataset Description

- **Homepage:** [Eureka5](http://www.eureka5.it)
- **Repository:** [Github](https://github.com/gsarti/verbalized-rebus)
- **Paper:** [Arxiv](https://arxiv.org/abs/2408.00584)
- **Contact:** [Gabriele Sarti](mailto:gabriele.sarti996@gmail.com)

> Last data update: September 17th, 2024. Refer to the [changelog](./CHANGELOG.md) for a list of revisions that can be loaded with the `revision` parameter in `load_dataset`.

### Dataset Summary

This dataset contains the original collection of over 200k first passes and solution for Italian rebuses published in various Italian magazines dating back to 1869. The original data are hosted in the [Eureka5](http://www.eureka5.it) platform of the [Associazione Culturale "Biblioteca Enigmistica Italiana - G. Panini"](http://www.enignet.it). This release includes a subset of verbalized rebuses created for the paper [Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses](https://arxiv.org/abs/2408.00584) by combining the original collection with Italian crossword definitions sourced from the [ITA_CW](https://huggingface.co/datasets/Kamyar-zeinalipour/ITA_CW) dataset.

For further information, see our [code](https://github.com/gsarti/verbalized-rebus) and [artifacts collection](https://huggingface.co/collections/gsarti/verbalized-rebus-clic-it-2024-66ab8f11cb04e68bdf4fb028).


### Data Splits

| Name | Number of Samples | Description
| --- | --- | --- |
| rebus | 223,007 | Transcribed rebus puzzles downloaded as-is from Eureka5 |
| verbalized_rebus | 83,158 | Subset of rebus puzzles with verbalized first passes using ItaCW definitions |
| llm_sft | 83,158 | Training and test splits of verbalized rebuses converted in chat format for fine-tuning LLMs |

### Rebus

#### Fields

- `FRASE`: Rebus solution phrase.
- `PRIMALET`: First pass of the rebus, from which the solution is derived via re-segmentation.
- `AUTORE`: Author of the rebus.
- `RIVISTA`: Magazine where the rebus was published.
- `MESE`: Month or issue number of the magazine.
- `ANNO`: Year of publication.
- `TIPO`: Rebus category (regular if empty). See [all available categories](http://www.eureka5.it/public/Leg_RebusSearchCrud).
- `NOTE`: Additional notes on the rebus.

#### Rebus Data Sample

```json
{
  "FRASE": "Non vi sono difficoltà a l'amore sincero",
  "PRIMALET": "N O N - viso - nodi - F fico - L T A A L - amo RE - S in cero",
  "AUTORE": "-",
  "RIVISTA": "RPC",
  "MESE": "0",
  "ANNO": "1869",
  "TIPO": "",
  "NOTE": ""
}
```

### Verbalized Rebus

#### Additional Fields

- `WORDS_PRIMALET`: Ordered list of words in the rebus first pass phrase (`PRIMALET`), whitespace-separated.
- `LETTERS_PRIMALET`: Ordered list of letters in the rebus first pass phrase (`PRIMALET`), whitespace-separated.
- `DIAGRAMMA_FRASE`: Solution key for the rebus solution (`FRASE`), derived automatically from word lengths.
- `FRASE_SEPARATED`: Whitespace-tokenized solution phrase (`FRASE`).
- `VERBALIZED_PRIMALET`: Verbalized first pass phrase (`PRIMALET`) using Italian crossword definitions.
- `VERBALIZED_PRIMALET_WITH_LEN`: Verbalized first pass phrase (`PRIMALET`) using Italian crossword definitions, including solution words lengths.

#### Verbalized Rebus Data Sample

```json
{
  "FRASE": "Non vi sono difficoltà a l'amore sincero",
  "PRIMALET": "N O N - viso - nodi - F fico - L T A A L - amo RE - S in cero",
  "AUTORE": "-",
  "RIVISTA": "RPC",
  "MESE": "0",
  "ANNO": "1869",
  "TIPO": "",
  "NOTE": "",
  "WORDS_PRIMALET": "viso nodi fico amo in cero",
  "LETTERS_PRIMALET": "N O N F L T A A L R E S",
  "DIAGRAMMA_FRASE": "3 2 4 10 1 1 ' 5 7",
  "FRASE_SEPARATED": "Non vi sono difficoltà a l ' amore sincero",
  "VERBALIZED_PRIMALET": "N O N [Volto, faccia] [Possono essere scorsoi] F [C'è anche quello d'India] L T A A L [Un' insidia subacquea] RE S [Alla moda... inglese] [Si accende in chiesa]",
  "VERBALIZED_PRIMALET_WITH_LEN": "N O N [Volto, faccia (4)] [Possono essere scorsoi (4)] F [C'è anche quello d'India (4)] L T A A L [Un' insidia subacquea (3)] RE S [Alla moda... inglese (2)] [Si accende in chiesa (4)]"
}
```

### Licensing and Contact Information

Please refer to our [LICENSE](LICENSE) file for details on the licensing of the data. For any questions or requests regarding the original data, please contact the [Eureka5 maintainers](mailto:bei.eureka5@gmail.com). For questions or requests regarding the verbalized rebuses or this repository, please contact [Gabriele Sarti](mailto:gabriele.sarti996@gmail.com)

### Citation Information

If you use this dataset in your work, please cite our paper as follows:

```bibtex
@article{sarti-etal-2024-rebus,
    title = "Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses",
    author = "Sarti, Gabriele and Caselli, Tommaso and Nissim, Malvina and Bisazza, Arianna",
    journal = "ArXiv",
    month = jul,
    year = "2024",
    volume = {abs/2408.00584},
    url = {https://arxiv.org/abs/2408.00584},
}
```

## Acknowledgements

We are grateful to the [Associazione Culturale "Biblioteca Enigmistica Italiana - G. Panini"](http://www.enignet.it/home) for making its rebus collection freely accessible on the [Eureka5 platform](http://www.eureka5.it).