Datasets:

techiaith
/

toxigen-data

annotations_creators:
  - no-annotation
language_creators:
  - machine-generated
language:
  - en
  - cy
multilinguality: translation
size_categories:
  - 100K<n<1M
source_datasets:
  - extended|toxigen/toxigen-data
task_categories:
  - text-classification
task_ids:
  - hate-speech-detection
pretty_name: toxigen-data
config_names:
  - annotated
configs:
  - config_name: annotated
    data_files:
      - split: test
        path: annotated/test-*
dataset_info:
  config_name: annotated
  features:
    - name: text
      dtype: string
    - name: text_cy_translation
      dtype: string
    - name: target_group
      dtype: string
    - name: factual?
      dtype: string
    - name: ingroup_effect
      dtype: string
    - name: lewd
      dtype: string
    - name: framing
      dtype: string
    - name: predicted_group
      dtype: string
    - name: stereotyping
      dtype: string
    - name: intent
      dtype: float64
    - name: toxicity_ai
      dtype: float64
    - name: toxicity_human
      dtype: float64
    - name: predicted_author
      dtype: string
    - name: actual_method
      dtype: string
  splits:
    - name: test
      num_bytes: 467413
      num_examples: 940
  download_size: 144090
  dataset_size: 467413

Cerdyn Data — `toxigen-data` (Cymraeg)

Cyfieithiad Cymraeg o set werthuso ToxiGen, set ddata sy'n cael ei defnyddio i archwilio sut mae Modelau Iaith Mawr (LLMs) yn ymdrin â iaith casineb ymhlyg ac iaith ragfarnllyd. Mae'r set wreiddiol, a grëwyd gan ymchwilwyr o Microsoft Research, MIT, AI2 a Phrifysgol Washington (Hartvigsen et al., ACL 2022), yn cynnwys datganiadau gwenwynig a niwtral wedi'u cynhyrchu gan beiriant am 13 grŵp lleiafrifol, gan ganolbwyntio ar gasineb cynnil nad yw'n dibynnu ar regfeydd na sarhad amlwg.

Mae'r fersiwn Gymraeg hon, a baratowyd gan Uned Technolegau Iaith Prifysgol Bangor (techiaith), yn galluogi gwerthuso modelau iaith Cymraeg yn erbyn yr un meincnodau diogelwch a thegwch ag a ddefnyddir ar gyfer y Saesneg, gan helpu i sicrhau nad yw datblygu LLM Cymraeg yn gadael bylchau yn y gwaith o liniaru ar niwed ieithyddol.

Rhybudd cynnwys: O reidrwydd, mae'r set ddata'n cynnwys enghreifftiau o iaith casineb a deunydd a allai beri tramgwydd. Bwriedir hi at ddibenion ymchwil yn unig.

Dataset Card — `toxigen-data` (Welsh)

A Welsh translation of the ToxiGen evaluation set, used to probe how Large Language Models (LLMs) handle implicit hate speech and biased language. The original dataset, created by researchers from Microsoft Research, MIT, AI2 and the University of Washington (Hartvigsen et al., ACL 2022), comprises machine-generated toxic and benign statements about 13 minority groups, focused on subtle hate that does not rely on slurs or overt profanity.

This Welsh version, prepared by the Language Technologies Unit at Bangor University (techiaith), allows Welsh-language models to be assessed against the same safety and fairness benchmarks used for English, helping to ensure that Welsh LLM development does not leave gaps in the work of mitigating linguistic harm.

Content warning: By its nature, the dataset contains examples of hate speech and material that may cause offence. It is intended for research use only.

Cerdyn Data — toxigen-data (Cymraeg)

Dataset Card — toxigen-data (Welsh)

Cerdyn Data — `toxigen-data` (Cymraeg)

Dataset Card — `toxigen-data` (Welsh)