Create dataset loader for Vietnamese Hate and Offensive Spans Detection (ViHOS)

[SamuelCahyawijaya]

Dataloader name: vihos/vihos.py
DataCatalogue: http://seacrowd.github.io/seacrowd-catalogue/card.html?vihos

Dataset	vihos
Description	This dataset consists of human-annotated hateful and offensive spans in Vietnamese Facebook and Youtube comments. Each comment has a corresponding list of indices indicating the characters included in these hate and offensive spans. Individual words and syllables are also tagged as inside or outside spans using the Inside-Outside-Beginning (IOB) tagging representation.
Subsets	-
Languages	vie
Tasks	Hate Speech Detection
License	MIT (mit)
Homepage	https://github.com/phusroyal/ViHOS
HF URL	-
Paper URL	https://aclanthology.org/2023.eacl-main.47

[elyanah-aco]

#self-assign

[github-actions]

Hi, may I know if you are still working on this issue? Please let @holylovenia @SamuelCahyawijaya @sabilmakbar know if you need any help.

[elyanah-aco]

@holylovenia @SamuelCahyawijaya @sabilmakbar

Was planning to implement ABUSIVE_LANGUAGE_PREDICTION task here.

Would like your thoughts on whether dataset can also support SPAN_BASED_ABSA task or not. Spans are BIO-tagged, but all spans would be labelled "offensive".

[github-actions]

Hi @, may I know if you are still working on this issue? Please let @holylovenia @SamuelCahyawijaya @sabilmakbar know if you need any help.

[holylovenia]

Spans are BIO-tagged

Hi @elyanah-aco, sorry I missed your question. Doesn't the sequence labeling version of the data use B- and I- for offensive words and O for others? This is my assumption based on a quick look of the data.