Dataset Card for tw-dictionary
本資料集整合中華民國公開之三部繁體中文辭典/成語典:
- dictionary-of-chinese-idioms:成語典
- mandarin-chinese-mini-dictionary:國語辭典簡編本
- revised-mandarin-chinese-dictionary:重編國語辭典修訂本
可作為繁體中文模型在用詞、成語、字義、注音等基礎語言知識上的補強語料。
Dataset Details
Dataset Description
這幾部辭典/成語典是繁體中文最具權威性的公開字/詞知識庫之一,內容涵蓋字音、字義、詞性、例句、出處典故等。本資料集將三部辭典分別作為獨立 config,方便依需求單獨使用或合併。
可用於:
- 增強模型對繁中字詞、成語、典故的覆蓋。
- 訓練字音/注音相關任務(部分 config 含注音資訊)。
- 作為文言/古文 / 成語使用情境之教學素材。
- Curated by: Huang Liang Hsun
- Language(s) (NLP): Traditional Chinese
- License: cc-by-sa-4.0(沿用原始公開授權條款)
Dataset Sources
- Repository: lianghsun/tw-dictionary
- Source: 中華民國公開辭典/成語典資料
- Paper: TBA
Uses
Direct Use
- 補強繁中模型對字音、字義、成語典故的知識覆蓋。
- 作為國文教學、語文 chatbot 之 RAG 知識庫。
- 衍生「字義解釋」「成語應用」「注音標註」等下游任務之資料。
Out-of-Scope Use
- 不適用作為法律、考試標準答案;雖然原辭典具權威性,下游使用時仍應對照原始官方頁面。
- 商用前請確認原始開放授權條款是否適用於目標應用。
Dataset Structure
三個 config 各對應一部辭典,內容欄位以條目為單位,視來源略有差異(字、詞、注音、字義、例句、出處等)。實際欄位以資料集載入後 features 為準。
Dataset Creation
Curation Rationale
通用模型在繁中字詞層面的精準度常受限:注音標錯、成語使用情境誤判、字義解釋過於簡化。本辭典資料是公認的權威語言資源,把它整理為易於訓練的格式,能直接提升模型在這些任務上的表現。
Source Data
Data Collection and Processing
- 從公開來源取得三部辭典/成語典資料。
- 解析條目欄位,拆分為 jsonl。
- 依辭典分目錄存放並建立對應 config。
Who are the source data producers?
原始辭典由中華民國主管機關編纂並公開。
Annotations
Annotation process
無新增人工標註;條目即為原始辭典內容。
Who are the annotators?
原始條目由原辭典編纂團隊撰寫。
Personal and Sensitive Information
辭典條目不含個人識別資料。
Bias, Risks, and Limitations
- 辭典反映特定編纂時期的用法,部分新詞、網路用語可能未涵蓋。
- 三部辭典的覆蓋面與顆粒度不同,下游使用時可能需要再做欄位對齊。
Recommendations
建議與其他現代語料一同使用,避免模型對「正規/典雅用詞」過度偏好而失去日常表達能力。
Citation
@misc{tw_dictionary,
title = {tw-dictionary: Traditional Chinese Dictionary and Idiom Reference Corpus},
author = {Huang, Liang Hsun},
year = {2025},
howpublished = {\url{https://huggingface.co/datasets/lianghsun/tw-dictionary}},
note = {Sourced from publicly available Traditional Chinese dictionaries.}
}
Dataset Card Authors
Dataset Card Contact
- Downloads last month
- 6