finepdfs-zhtw / README.md
thliang01's picture
docs: update dataset summary and fix typos
315e0b0 verified
metadata
license: mit
task_categories:
  - text-generation
language:
  - zh
  - en
tags:
  - zh-tw
  - Taiwan
  - R.O.C
  - pdf
  - finepdfs
  - ocr
pretty_name: A Traditional Chinese PDF–Text Dataset for LLM training
size_categories:
  - 10M<n<100M

Dataset Card for finepdfs-zhtw (WIP)

FinePDFs-zhtw-v2.3

Brief Summary (Preliminary)

This section describes the initial design intent of the project. It serves as a first-edition summary and may be revised or expanded as the dataset scope, governance, and contribution process become more clearly defined.

本計畫旨在建立一套以繁體中文為核心、可合法使用、可長期維護的 PDF 文本資料集,作為繁體中文語言模型在 文件理解(Document Understanding)、OCR 後處理與微調訓練(fine-tuning) 等任務上的基礎資料來源。

整體設計將參考 Hugging Face 社群所建立之 finepdf 類型資料集的精神與結構,但不直接複製其資料來源或內容,而是打造一個專為繁體中文情境設計、由台灣社群自行貢獻、具備清楚授權邊界的版本。

如何幫忙本專案

本專案正在進行中,歡迎任何伙伴加入幫忙,一起讓繁體中文的預訓練語料更多元及完善。