|
--- |
|
annotations_creators: |
|
- crowdsourced |
|
language_creators: |
|
- found |
|
language: |
|
- en |
|
license: |
|
- other |
|
multilinguality: |
|
- monolingual |
|
pretty_name: RiddleSense |
|
size_categories: |
|
- 1K<n<10K |
|
source_datasets: |
|
- riddle_sense |
|
tags: |
|
- riddles |
|
- riddler |
|
- non-commercial |
|
--- |
|
# _RiddleSense_ ++: Evaluating LLMs' Riddling abilities |
|
|
|
Some cleaning and other modifications to make [the riddle_sense dataset](https://huggingface.co/datasets/riddle_sense) more suitable for training language models to generate riddles via standard `text generation` |
|
|
|
## Notable changes |
|
|
|
- reformatting to use special tags indicating the question and answer components |
|
- normalization: whitespace normalization, etc with `clean-text` |
|
- **spell correction** the original dataset has numerous spelling errors; these are fixed using `BertChecker` from the `neuspell` library **for the questions only**. The answers were left as-is as there is some specific terminology/puns that get changed by SC. |
|
|
|
## Notes & Gotchas |
|
|
|
- the original dataset has a non-commercial clause, so this does too. |
|
|
