Datasets:

msislam
/

marc-code-mixed-encodings

dataset_info:
  features:
    - name: input_ids
      sequence: int64
    - name: attention_mask
      sequence: int64
    - name: labels
      sequence: int64
  splits:
    - name: train
      num_bytes: 245408208
      num_examples: 60000
    - name: test
      num_bytes: 20286240
      num_examples: 5000
    - name: validation
      num_bytes: 20286240
      num_examples: 5000
  download_size: 27474371
  dataset_size: 285980688

Dataset Card for "marc-multilingual-encodings-v4"

marc-code-mixed-small

These encodings are based on the review_tokens of msislam/marc-code-mixed-small.

It contains German (DE), English (EN), Spanish (ES), and French (FR) languages.

The labels are 0 (DE), 1 (EN), 2 (ES), and 3 (FR).

Each review contains all four languages.

Total number of tokens:

In training set: 10195342
In test set: 842760
In validation set: 842760

The encodings are created using the tokenizer xlm-roberta-base.