README.md

metadata

license: other
task_categories:
  - text-classification
language:
  - en
  - hi
  - bn
  - gu
  - kn
  - ml
  - mr
  - or
  - ta
  - te
tags:
  - language-switch-detection
  - language-identification
  - indic
  - evaluation
  - benchmark
pretty_name: LSD Evaluation Dataset
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: language_switch
      dtype: string
    - name: text
      dtype: string
    - name: script_type
      dtype: string
    - name: language
      dtype: string
  splits:
    - name: train
      num_bytes: 756376
      num_examples: 10048
  download_size: 216426
  dataset_size: 756376

blue-machines/lsd_evaluation_dataset

Held-out evaluation set for language-switch detection / LID (LSD_final_dataset.csv).

Columns

• language_switch
• text
• script_type
• language

Stats

• Rows: 10,048
• Languages: Bengali, English, Gujarati, Hindi, Kannada, Malayalam, Marathi, Oria, Tamil, Telugu
• Switch labels: Bengali, English, Gujarati, Hindi, Kannada, Malayalam, Marathi, Oria, Tamil, Telugu, no_switch

Load

from datasets import load_dataset
ds = load_dataset("blue-machines/lsd_evaluation_dataset")
print(ds["train"][0])