README.md

---
language:
- id
- jv
- su
license: cc-by-4.0
task_categories:
- text-classification
task_ids:
- hate-speech-detection
tags:
- code-mixing
- hate-speech
- Indonesian
- Javanese
- Sundanese
- augmentation
- LLM
pretty_name: Code-mixed Indonesian (Javanese & Sundanese) Hate Speech Dataset
configs:
- config_name: Indonesian-Javanese
  data_files:
  - split: train
    path: Indonesian-Javanese/train-*
  - split: test
    path: Indonesian-Javanese/test-*
- config_name: Indonesian-Sundanese
  data_files:
  - split: train
    path: Indonesian-Sundanese/train-*
  - split: test
    path: Indonesian-Sundanese/test-*
dataset_info:
- config_name: Indonesian-Javanese
  features:
  - name: original
    dtype: string
  - name: augmented_1
    dtype: string
  - name: augmented_2
    dtype: string
  - name: augmented_3
    dtype: string
  - name: augmented_4
    dtype: string
  - name: augmented_5
    dtype: string
  - name: label
    dtype: string
  - name: llm
    dtype: string
  - name: strategy
    dtype: string
  - name: language_pair
    dtype: string
  - name: split
    dtype: string
  splits:
  - name: train
    num_bytes: 36494185
    num_examples: 45144
  - name: test
    num_bytes: 196788
    num_examples: 1000
  download_size: 23760418
  dataset_size: 36690973
- config_name: Indonesian-Sundanese
  features:
  - name: original
    dtype: string
  - name: augmented_1
    dtype: string
  - name: augmented_2
    dtype: string
  - name: augmented_3
    dtype: string
  - name: augmented_4
    dtype: string
  - name: augmented_5
    dtype: string
  - name: label
    dtype: string
  - name: llm
    dtype: string
  - name: strategy
    dtype: string
  - name: language_pair
    dtype: string
  - name: split
    dtype: string
  splits:
  - name: train
    num_bytes: 36416999
    num_examples: 45144
  - name: test
    num_bytes: 196468
    num_examples: 1000
  download_size: 23493088
  dataset_size: 36613467
---

# Code-mixed Indonesian Hate Speech Dataset

Manually annotated hate speech dataset for **Indonesian-Javanese** and
**Indonesian-Sundanese** code-mixed text, enriched with LLM-generated augmentations.