Using the API

Languages

A card has one pronunciation field and one example. Most languages write them the same way; the languages on this page each have a format of their own, and Lymi's AI writes to it when it enriches a card.

Every other language

pronunciation holds IPA or a plain respelling, and example is a plain sentence. Two cards are duplicates when their term and language match.

Japanese

Tag ja, written in Jpan.

pronunciation, called Reading in the app: The reading in hiragana. Leave it out when the term is written only in kana. 食べる → たべる
example: The reading in hiragana in square brackets right after each run of kanji, and nothing over kana. 毎朝[まいあさ]パンを食[た]べます。 Each reading matches \[([\p{Script=Hiragana}\p{Script=Katakana}ー]{1,30})\] with the u flag.
Duplicates: The term and language, unless both cards have a reading and the readings differ: then they are two words.

Chinese

Tag zh, written in Hans or Hant.

pronunciation, called Reading in the app: Pinyin in tone marks and citation tones, as dictionaries write them: 不是 bù shì and 一起 yīqǐ, never the spoken bú or yì. A neutral tone has no mark (péngyou, dōngxi), an erhua r joins the syllable before it (yīdiǎnr), and letters stay as written (kǎlā OK). Where a character has two readings, as 行 háng and xíng do, the meaning decides. Not IPA or tone numbers. 学习 → xuéxí
example: Pinyin in square brackets right after each word of every example, one syllable for each character, in tone marks, and nothing after punctuation, letters or digits. 我[wǒ]去[qù]银行[yínháng]取钱[qǔqián]。 Each reading matches \[([\p{Script=Latin}\p{M}0-5 '’:]{1,60})\] with the u flag.
Duplicates: The term and language, unless both cards have a reading and the readings differ: then they are two words.

Ukrainian

Tag uk, written in Cyrl.

pronunciation, called Stress in the app: The term with U+0301 over the stressed vowel of each word of two or more syllables. Leave it out when every word of the term has one syllable. Not IPA. замок → за́мок
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, unless both cards have a reading and the readings differ: then they are two words.

Russian

Tag ru, written in Cyrl.

pronunciation, called Stress in the app: The term with U+0301 over the stressed vowel of each word of two or more syllables, and ё where the term writes е. Leave it out when every word of the term has one syllable or is written with ё, and none writes е for ё. Not IPA. замок → за́мок
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, unless both cards have a reading and the readings differ: then they are two words.

Tamil

Tag ta, written in Taml.

pronunciation, called Reading in the app: Only for a word not said as spelled, mostly a loanword, the reading in Latin letters: long vowels with a macron (ā ī ū ē ō), retroflex consonants with a dot below (ṭ ḍ ṇ ḷ), ழ as zh, ஷ as sh, and each stop voiced where it is said voiced. A word in grantha letters (ஜ ஷ ஸ ஹ) is said as spelled too. Leave it out when the word is said as spelled, because Lymi reads the spelling itself. Not IPA, and not the source word's own spelling: bēg for பேக், never bag. பஸ் → bas
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Cantonese

Tag yue, written in Hant or Hans.

pronunciation, called Reading in the app: Jyutping with a tone number from 1 to 6 after each syllable, as dictionaries write it: 你好 nei5 hou2, 唔該 m4 goi1. Write a changed tone as it is said (仔女 zai2 neoi2), letters as they are said or as written (T恤 ti1 seot1, 卡拉OK kaa1 laa1 OK), and where a character has two readings, as 行 hong4 and hang4 do, let the meaning decide. Not Yale, pinyin or IPA, and never tone marks. 學習 → hok6 zaap6
example: Jyutping with tone numbers in square brackets right after each word of every example, one syllable for each character, and nothing after punctuation, letters or digits. 我[ngo5]去[heoi3]銀行[ngan4 hong4]攞錢[lo2 cin2]。 Each reading matches \[([a-zA-Zａ-ｚＡ-Ｚ1-6¹²³⁴⁵⁶１-６ '’-]{1,60})\] with the u flag.
Duplicates: The term and language, unless both cards have a reading and the readings differ: then they are two words.

Korean

Tag ko, written in Kore.

pronunciation, called Pronunciation in the app: How the term is said, in Hangul, only where that differs from its spelling. Leave it out when the term is said as written. Not romanization or IPA. 학교 → 학꾜
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Filipino

Tag fil, written in Latn.

pronunciation, called Pronunciation in the app: The term with its stress marks: an acute on the stressed vowel (báhay), a grave on a final vowel that ends in a glottal stop (bátà), and a circumflex where the two meet (basâ). Not IPA. bata → bátà
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Vietnamese

Tag vi, written in Latn.

pronunciation, called Pronunciation in the app: How the term is said, in Vietnamese spelling with every tone mark, only where the term does not show it: a term typed without its tone marks, or an abbreviation (km → ki-lô-mét). Leave it out when the term is written with its tone marks. Not IPA. cam on → cảm ơn
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Georgian

Tag ka, written in Geor.

pronunciation, called Pronunciation in the app: The term romanized letter by letter in the national system, with an apostrophe after each ejective: კ k', პ p', ტ t', წ ts', ჭ ch', ყ q'. Not IPA. ჭიქა → ch'ika
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Armenian

Tag hy, written in Armn.

pronunciation, called Pronunciation in the app: How the term is said in Eastern Armenian, romanized: an apostrophe after each aspirated consonant (թ t', փ p', ք k', ց ts', չ ch'), ու as u, ը as ə, and ե as ye, ո as vo and և as yev at the start of a word. Not IPA. քույր → k'uyr
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Lithuanian

Tag lt, written in Latn.

pronunciation, called Pronunciation in the app: The term with its accent mark, as Lithuanian dictionaries print it: an acute, grave or tilde on the stressed syllable (výras, kalbà). Not IPA. namas → nãmas
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Danish

Tag da, written in Latn.

pronunciation, called Pronunciation in the app: How the word is said, in IPA as Den Danske Ordbog writes it: stress ˈ, length ː and stød ˀ, leaving out the letters that are not said. Not a respelling. hus → ˈhuˀs
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

French

Tag fr, written in Latn.

pronunciation, called Pronunciation in the app: IPA, without slashes. In a phrase, mark each liaison with ‿ (les amis → lez‿ami). Not a respelling. oiseau → wazo
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Portuguese

Tag pt, written in Latn.

pronunciation, called Pronunciation in the app: IPA as the word is said in Portugal, without slashes; a pt-BR deck is said as in Brazil. Not a respelling. obrigado → ɔbɾiˈɣaðu
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Brazilian Portuguese

Tag pt-BR, written in Latn.

pronunciation, called Pronunciation in the app: IPA as the word is said in Brazil, without slashes. Not a respelling. obrigado → obɾiˈɡadu
example: A plain sentence, with no readings inside it.
Duplicates: The term and language, as for any language.

Ask Lymi

The same facts come from GET /api/languages, which needs no key and may be cached for an hour. An assistant gets them from list_decks, as a languages map for the languages of the learner's decks.

Terminal

curl "$LYMI_URL/api/languages"

Response · 200, first entry

{
  "languages": [
    {
      "tag": "ja",
      "name": "Japanese",
      "scripts": [
        "Jpan"
      ],
      "pronunciation": {
        "label": "reading",
        "format": "The reading in hiragana. Leave it out when the term is written only in kana.",
        "sample": {
          "term": "食べる",
          "pronunciation": "たべる"
        }
      },
      "example": {
        "format": "The reading in hiragana in square brackets right after each run of kanji, and nothing over kana.",
        "pattern": "\\[([\\p{Script=Hiragana}\\p{Script=Katakana}ー]{1,30})\\]",
        "sample": "毎朝[まいあさ]パンを食[た]べます。"
      },
      "duplicates": "term-and-reading"
    }
  ]
}