el2e10's picture
Update README.md
1dcf30b verified
metadata
language:
  - hi
license: cc
size_categories:
  - 1K<n<10K
task_categories:
  - conversational
dataset_info:
  features:
    - name: inputs
      dtype: string
    - name: targets
      dtype: string
    - name: template_lang
      dtype: string
    - name: template_id
      dtype: int64
  splits:
    - name: train
      num_bytes: 634314
      num_examples: 1156
  download_size: 289336
  dataset_size: 634314
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Description

This dataset is derived from the already existing dataset made by AI4Bharat. We have used the IndicSentiment dataset of AI4Bharat to create an instruction style dataset.

IndicSentiment is a multilingual parallel dataset for sentiment analysis. It encompasses product reviews, translations into Indic languages, sentiment labels, and more. The original dataset(IndicSentiment) was made available under the cc-0 license.

This dataset comprises a singular split (train) and encompasses over 1150+ rows.

Template

The following template was used for converting the original dataset:

#Template 1
prompt:
Translate from English to {target_language}:
{ENGLISH_REVIW}

completion:
{INDIC_REVIEW}
#Template 2
prompt:
Translate this sentence to {target_language}:
{ENGLISH_REVIW}

completion:
{INDIC_REVIEW}
#Template 3
prompt:
What's the {target_language} translation of this language:
{ENGLISH_REVIW}

completion:
{INDIC_REVIEW}
#Template 4
prompt:
Can you translate this text to {target_language}:
{ENGLISH_REVIW}

completion:
{INDIC_REVIEW}