Definitions
CC-CEDICT, published by MDBG: 124,725 entries from the 23 July 2026 snapshot used by the app. Simplified and traditional spellings, pinyin and English definitions are shared under CC BY-SA 4.0. This is a dated snapshot, not a live feed of the latest dictionary edits.
We split articles into definition lines and built search indexes. The exported dictionary data keeps the same licence. Original project · Attribution
CFDICT: 58,051 entries have French text from the app’s 8 August 2026 source, under CC BY-SA 3.0. Source matching groups some pronunciations, so French definitions may cover more than the selected English sense. Entries that only repeated the English fallback were omitted. Original project · Attribution
Search follows the app’s native algorithm: Chinese prefixes with exact spelling first, mixed Chinese–pinyin lookup, pinyin completion and numbered-tone filtering, phrase and token ranking, and spelling-variant collapsing. Ambiguous Latin queries offer separate meaning and pinyin results. Ukrainian follows its own short-token, connecting-word and BM25 ranking rules. Jieba frequency orders the native candidate pools. Jieba licence
Ukrainian meanings and examples
The same app-admitted Ukrainian corpus supplies 121,420 dictionary entries with 193,711 meaning segments. The adaptation by Dmytro Polianskyi’s Ukrainian CEDICT project retains CC BY-SA 4.0. These translations were reviewed with models; they have not been validated by a human. Missing translations are labelled as English fallback.
There are 1,139 licensed Ukrainian sentence translations from Tatoeba. Each keeps its contributor credits and original source links. Examples without Ukrainian text retain a clearly labelled English translation.
Ukrainian attribution and adaptation notice · Native search and Ukrainian data manifest
Sentence examples
76,255 Chinese–English sentence pairs from Tatoeba, snapshot 18 July 2026, under CC BY 2.0 FR. Each example links to its original Chinese and English sentences and retains the contributor credit supplied with the source. Some source records give the generic credit “Tatoeba contributor”.
The app’s import converted Chinese to simplified characters, normalised English text and matched words automatically. These are community contributions, not a human-reviewed course. Word links may open a different sense or pronunciation; use the surrounding sentence to check the meaning. We do not invent missing sentence pinyin.
Examples are available for 19,674 dictionary entries. An entry without a linked pair says so. We do not republish the app’s legacy example collection without separate rights evidence. Example attribution and adaptation notice
Character animations
Make Me a Hanzi: 9,574 characters and 112,617 ordered strokes. The graphics derive from Arphic fonts, Copyright © 1999 Arphic Technology Co., Ltd., and are distributed under the Arphic Public License.
We use the original SVG outlines and median points from the pinned upstream release, split into downloadable JSON files on 21 October 2026. Each file carries its licence and modification notice. The modified graphical data is freely available under the same licence. The page’s interface code is separate from the graphical data.
Coverage includes common simplified and traditional characters; some rare characters have no animation. Stroke order follows the source’s PRC convention. Regional writing conventions may differ.
Pinned source release · Original copying notice · Graphics files and hashes
Word and character reference from the app
FMLD supplies 96,273 synonym, antonym and measure-word links under CC BY-SA 4.0. Chinese Open Wordnet 2.0 and Princeton WordNet 3.0 supply 87,118 concept links under their WordNet licences. These are headword connections, not guarantees that every sense is interchangeable.
Make Me a Hanzi supplies 9,574 character records: decomposition, radicals and explanatory hints, under LGPL 3.0 or later. We provide the unmodified source, copying notice and full licence alongside the transformed JSON. Unicode Unihan 17.0 supplies 102,998 character references under Unicode License v3; Hong Kong school grades retain their regional meaning.
The Leipzig cmn-tw_web_2020_30K corpus supplies 24,486 word-frequency records and 14,003 immediate-neighbour collocations under Creative Commons Attribution. These statistics describe this dated Taiwanese web sample. We do not republish its source web sentences.
Extra examples include 4,503 Tatoeba French pairs (CC BY 2.0 FR), 472 Chinese Dialogues speaker turns by Wen-Hua Teng / COERLL, The University of Texas at Austin (CC BY 4.0), and 1,999 Google FLEURS Chinese examples (CC BY 4.0). Each displayed example retains its source, creator credit and licence. App-generated pinyin and automated word matching are adaptations.
Recorded pronunciation comes from audio-cmn / Shtooka, creators Yue Tan and Chen Wang, curated by Hugo Lopez, PLIDAM, INALCO, under CC BY-SA 3.0 US. The app’s pinned source is ff9ed3d0c631195bd2c06f39450f3264c7124040. Original MP3 clips were regrouped into browser packs without changing the audio. Original audio project · Audio licence.
All exported files and hashes · Source versions and changes · Attribution and licence notices · Chinese Open Wordnet licence · Princeton WordNet licence · FMLD source.
Download and reuse the data
The same files used by the dictionary are public. The manifests list every data file, its size, its hash and its source. Dictionary entries, examples and stroke graphics retain their separate licences and attribution requirements.
Search and saved words
Searches run in the browser using small files from this website. Your typed query is not submitted to a search service. Entry links and queries use the part of the address after #, which the browser may retain in its history.
Saved vocabulary stays in this browser. You can remove words, clear the list, or download it to import into Laowai. Use web study for custom lists, eight practice skills, daily reviews and per-skill progress. Optional iCloud sync shares lists and word progress with Laowai on the same Apple ID. Browser Mandarin speech is synthesised; a separately labelled button plays an attributed recording when one is available. Recordings can use a different reading of a headword; use the pinyin to check polyphonic words.