README.md

metadata

license: odc-by
language:
  - zh
  - yue
task_categories:
  - text-generation
size_categories:
  - 1M<n<10M
pretty_name: FinePDFs-zh
configs:
  - config_name: yue
    data_files:
      - split: train
        path: data/yue/train/*
      - split: test
        path: data/yue/test/*
  - config_name: zhcn
    data_files:
      - split: train
        path: data/zhcn/train/*
      - split: test
        path: data/zhcn/test/*
  - config_name: zhcn_classical
    data_files:
      - split: train
        path: data/zhcn_classical/train/*
      - split: test
        path: data/zhcn_classical/test/*
  - config_name: zhtw
    data_files:
      - split: train
        path: data/zhtw/train/*
      - split: test
        path: data/zhtw/test/*
  - config_name: zhtw_classical
    data_files:
      - split: train
        path: data/zhtw_classical/train/*
      - split: test
        path: data/zhtw_classical/test/*

FinePDFs-zh dataset card

Introduction

FinePDFs-zh is a fine-grained classified pdf dataset derived from the cmn_Hani subset of FinePDFs. Each sample is classified into Traditional Chinese, Simplified Chinese, Cantonese, Classical Chinese (Traditional), and Classical Chinese (Simplified) using MusubiAI/ZHLID.

Limitation

Data may be classified under the wrong label due to misclassification by the MusubiAI/ZHLID model. Users are encouraged to perform additional verification or processing before practical use.