Data mirror for iTaigi. This site provides direct access to the exported iTaigi crowdsourced dictionary data. The files are hosted on GitHub Pages for easy access.
版權聲明
- 程式授權：MIT License
- 辭典內容：CC0「不保留權利」授權
- iTaigi隱私權保護條款
- Version ID:
20261019-0110 - Last Updated: 2026-10-19
You can check manifest.json for the latest version information programmatically.
The files are organized by version. The latest files are in public/20261019-0110/.
| Field | Description |
|---|---|
id |
Unique id: content hash of the four source fields, see below |
華語 |
Mandarin Chinese gloss |
漢字 |
Taigi Han characters |
羅馬字 |
Taigi Roman Orthography in KPL (Kàu-io̍k Pō͘ Lô-má-jī, the MOE standard) |
來源 |
Contributor name or 匿名 (anonymous) |
link |
iTaigi dictionary page for the entry's 華語 |
The JSON file has a single key 資料 holding an array of entries with the fields above. The CSV has the same columns in the same order.
The iTaigi export carries no identifier of its own, so id is derived from the entry's content: the first 16 hex digits of the SHA-256 of its four source fields 華語, 漢字, 羅馬字 and 來源, each normalized to Unicode NFC and joined with U+001F (unit separator). An unchanged entry therefore keeps its id across versions, and any change to its text or contributor gives it a new id.
The export sometimes contains several rows that are identical in all four fields. Only the first is kept, so every id is unique; see Skipped Sources below.
link is the iTaigi dictionary page for the entry's 華語, for example https://itaigi.tw/k/%E8%A8%8E%E5%8E%AD/ for 討厭. iTaigi has a page per Mandarin headword, not per answer, so the page lists every Taigi answer for that headword and entries that share a 華語 share a link. The headword is percent-encoded in the URL.
iTaigi bundles entries imported from two other dictionaries. They are not crowdsourced iTaigi content and are available from their own sources, so this mirror drops every entry whose 來源 is one of:
來源 |
Entries in version 20261019-0110 |
|---|---|
台文華文線頂辭典 |
64,372 |
臺灣閩南語常用詞辭典 |
26,911 |
Rows identical to an earlier row in all four fields are skipped as well; version 20261019-0110 has 65 of them. Of the 111,592 entries in the source export, 20,244 remain after skipping. The raw export in tangloo/ is kept unchanged but is not committed to this repository.
辭典資料下載
Original data source: https://itaigi.tw/