---
dataset_info:
  features:
  - name: text
    dtype: string
  - name: label
    dtype:
      class_label:
        names:
          '0': negative
          '1': neutral
          '2': positive
  - name: source
    dtype: string
  - name: language
    dtype: string
  splits:
  - name: train
    num_bytes: 69063358
    num_examples: 270399
  - name: validation
    num_bytes: 2354116
    num_examples: 10857
  - name: test
    num_bytes: 2744290
    num_examples: 14465
  download_size: 38628976
  dataset_size: 74161764
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
source_datasets:
- tyqiangz/multilingual-sentiments
---

# multilingual-sentiments

Multilingual sentiments (tyqiangz): 3-way sentiment in 12 languages, merged from public corpora (`source`).

Original data: [tyqiangz/multilingual-sentiments](https://huggingface.co/datasets/tyqiangz/multilingual-sentiments). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/upload_repackaged.py](https://github.com/sileod/tasksource/blob/main/scripts/upload_repackaged.py).