1. Reading the transcription (Frellesvig–Whitman notation)
ONCOJ gives each poem in its original script (man'yōgana and logograms) side by side with a phonemic romanization. The romanization follows the notation of Frellesvig and Whitman (F&W), which writes the kō-rui / otsu-rui distinctions of Old Japanese (the so-called jōdai tokushu kanazukai) by inserting w or y between consonant and vowel.
Kō and otsu syllables
| vowel | kō (甲) | otsu (乙) | no distinction | consonants with the distinction (in the corpus) |
|---|---|---|---|---|
| i | Ci pi ‘sun’ | Cwi pwi ‘fire’ | Ci ti ‘blood’ | k g p b m |
| e | Cye mye ‘woman’ | Ce me ‘eye’ | Ce te ‘hand’ | k g p b m |
| o | Cwo kwo ‘child’ | Co ko ‘this’ | Co po ‘ear of grain’ | k g t d n s z y r m |
Note which side gets the extra letter: otsu for i (kwi = ki₂), kō for e (kye = ke₁) and kō for o (kwo = ko₁). Plain ki ke ko are therefore ki₁, ke₂, ko₂ (or syllables without the distinction).
| F&W | ki | kwi | kye | ke | kwo | ko |
|---|---|---|---|---|---|---|
| subscript notation | ki₁ | ki₂ | ke₁ | ke₂ | ko₁ | ko₂ |
The mwo / mo distinction is found only in the Kojiki songs (37 tokens of mwo in this corpus).
Vowel-initial syllables and other consonants
- Bare vowels are distinct from y- and w-syllables: i / wi, e / ye / we, o / wo. Here wi ye wo are syllables with the glides w/y, not kō/otsu markings.
- The h-row of later Japanese is p (pana ‘flower’); voiced obstruents are b d g z; the others are k s t n m y r w.
- w and y are a notational convention for writing the distinctions in ASCII, not necessarily a claim about pronunciation; the phonetic values of the kō/otsu vowels are debated.
- In the lexicon, bound morphemes are written with hyphens (-ki-, the past auxiliary), and each lemma has an ID such as L050083.
2. ONCOJ and Chūnagon (Corpus of Historical Japanese)
Besides ONCOJ, Old Japanese can be searched in the Nara period series of NINJAL's Corpus of Historical Japanese (CHJ) through the search application Chūnagon. Both involve NINJAL, but they are built differently and are good at different things.
| ONCOJ (the data on this page) | CHJ Nara series (Chūnagon) | |
|---|---|---|
| Texts | All Old Japanese poetry: Kojiki and Nihon shoki songs, Fudoki songs, Bussokuseki-ka, Man'yōshū, Kakyō hyōshiki, Shoku Nihongi songs, Jōgū Shōtoku hōō teisetsu; about 90,000 words | I Man'yōshū (c. 100,000 words), II Shoku Nihongi senmyō (c. 20,000), III Engi-shiki norito (c. 9,000); no Kojiki/Nihon shoki songs |
| Text | Original script and F&W phonemic transcription, with kō/otsu in the transcription itself; mode of writing (phonographic/logographic) per word | A kanji-kana reading text (for the Man'yōshū, from the Shōgakukan Shinpen Nihon koten bungaku zenshū) with word annotation; the original text can be consulted. Readings are in kana, so kō/otsu must be checked in the original man'yōgana |
| Word units | Lemmas of its own dictionary with English glosses; free words and bound morphemes (auxiliaries, inflectional endings) are separated | UniDic short and long units; lemma, POS and inflection follow the same scheme as the later-period subcorpora and modern corpora |
| Syntax | Constituency trees (following the Keyaki Treebank) with grammatical functions (subject, object, relative clause …) | None (search by word sequence and co-occurrence) |
| Access | ONCOJ website (tree search, dictionary) and this page; data downloadable, annotation CC BY 4.0 | Chūnagon (free, registration required); short/long-unit and string search with context conditions |
· Questions about structure (kakari-musubi, relative clauses, word order), searching by form including kō/otsu, or covering all Old Japanese poetry including the chronicle songs → ONCOJ
· Comparison with later periods on the same criteria, the prose of senmyō and norito, or vocabulary searches on a reading text → Chūnagon (CHJ)
The same poem may be read (kun) or segmented differently in the two corpora. The ONCOJ tree view links to the corresponding CHJ Nara-period text.
3. How this search page handles the data
- The data come from the ONCOJ psd (tree) files. Lemma-ID nodes are folded into their POS nodes and writing nodes (PHON, LOG …) become word attributes; the annotation itself is unchanged. See “Tags & structure” → “Data model” on the search page.
- “ignore kō/otsu” removes the w/y (kwo→ko, pwi→pi, kye→ke) before matching, in both the data and the query; “ignore voicing” treats g/k z/s d/t b/p as the same.
- In “Words” mode Old Japanese can also be typed in kana (ひとのこ). Kana cannot show kō/otsu or voicing, so such searches ignore both.
- Kakyō hyōshiki is not included in this version.
4. Sources and links
- Oxford-NINJAL Corpus of Old Japanese (ONCOJ): https://oncoj.ninjal.ac.jp/ (the transcription and the list of texts are on the front page)
- NINJAL, Corpus of Historical Japanese: https://clrd.ninjal.ac.jp/chj/; Chūnagon: https://chunagon.ninjal.ac.jp/
- Keyaki Treebank (the parsing principles followed by ONCOJ): http://www.compling.jp/keyaki/
Please report errors or unclear points in this page with the Feedback button on the search page.