Guide Home & Living 6 min read

Over a third of a Hokkien dictionary's syllables change between Pe̍h-ōe-jī and Tâi-lô

Pe̍h-ōe-jī (POJ) and Tâi-lô are two ways of writing Hokkien in Latin letters, and Tâi-lô is the scheme Taiwan's Ministry of Education announced in 2006. Respelled from one to the other, 37% of the syllables in the ministry's Hokkien dictionary change, mostly through a few fixed letter pairs and a tone mark that moves.

Eva Chin
Business & Chinese Culture Correspondent
Published 15 Oct 2026, 2:13 PM (SGT)
Share:
A temple hall with a curved red-tiled roof, carved dragons on the ridge and red lanterns at the doors A temple hall with a curved red-tiled roof, carved dragons on the ridge and red lanterns at the doors Photo by falco on Pixabay
Advertisement

We respelled every headword reading in Taiwan's Ministry of Education Hokkien dictionary from Tâi-lô into Pe̍h-ōe-jī (POJ), and 18,663 of its 50,105 headword syllables (37%) came out spelled differently. Of 24,594 headwords (the words entries are listed under) with a reading, 14,307 (58%) contain at least one changed syllable. We counted the first reading of each in the g0v copy of the ministry's dictionary data, last updated in May 2023. A headword list gives a rare word the same weight as a common one, so a count over running text could differ. Nearly all of the change comes from a short, fixed set of rules. POJ writes ch where Tâi-lô writes ts, and a dotted o where Tâi-lô doubles the o. The rest is a few vowel pairs, a nasal sign and a tone mark that sits on a different letter.

Ts and tsh shoulder nearly 40% of the changed syllables

A Hokkien syllable is described as an initial, the consonant it starts with, plus a final, everything after it. The ministry's handbook for Tâi-lô, 《臺灣閩南語羅馬字拼音方案使用手冊》 (user's manual for the Taiwan Southern Min romanisation scheme), has a cover dated March 2007. It counts 18 initials and sets them beside POJ in a comparison table. Two differ: Tâi-lô's ts and tsh appear in the POJ column as ch and chh. The manual's example sentence uses tsia̍h, which POJ spells chia̍h.

In the dictionary those two initials sit in 7,360 syllables, 4,506 with ts and 2,854 with tsh. By our count that is 39.4% of the 18,663 changed syllables. In 5,756 of them the initial is the only letter that moves. Hong Kong's place names raise a similar ts-or-ch question for Cantonese.

A doubled oo does the dotted o's job

POJ writes one vowel as an o with a small dot at its upper right, o͘. Tâi-lô writes it as plain oo. The ministry's own name shows it: Kàu-io̍k-pōo in Tâi-lô, Kàu-io̍k-pō͘ in POJ. The manual keeps the older letter as an option. Its note on the finals reads 「以 oo 為正式版,以 o·為傳統版」: oo is the formal version and o· the traditional one. Nasal vowels, sounded partly through the nose, get the same treatment, with nn formal and a raised ⁿ traditional. So khuànn in Tâi-lô is khòaⁿ in POJ, khoann in plain letters.

In our count the oo vowel changed 2,095 syllables and the nasal sign 3,109. Tâi-lô's ua became POJ's oa in 3,414, so Tâi-uân reads Tâi-oân. Ue became oe in 1,407 and ing became eng in 2,194, with ik to ek far behind. One syllable can carry two of these changes, so the counts overlap and do not add up.

Placement rules assign the same tone mark to a different vowel

Both systems write Hokkien's tones as accents over a vowel or as a digit, as in tai5. The manual takes POJ's traditional tone marks as Tâi-lô's formal scheme and allows numbers where marks are inconvenient. The manual ranks the vowels a, then oo, then e or o, then i or u. All 50,105 dictionary syllables carry the mark where that order and the manual's listed exceptions put it, in our check.

POJ follows six placement rules, which we know from Wikipedia's summary; it cites a 2001 source we did not open. In a two-letter oa or oe they put the mark on the o, so 我 ("I") is guá in Tâi-lô and góa in POJ. Wikipedia notes "some debate" over oa and oe. Embree's A Dictionary of Southern Min (1973), written in POJ and digitised by ChhoeTaigi, marks the o in 1,786 two-letter oa syllables and 2,005 oe syllables, and never the second vowel. Usually the mark moves because the letters under it change. In ui the letters agree and only the mark moves, from Tâi-lô uī to POJ ūi: 670 syllables, 3.6% of the changed ones.

Advertisement

Embree and the ministry handle the nasal before h in opposite orders

A checked syllable ends in p, t, k or h and takes tone 4 or tone 8. When a nasal vowel is checked by h, the ministry's POJ column writes the ⁿ after the h, as in ahⁿ and iahⁿ. Embree writes it before the h every time, as in saⁿh. We found 21 such tokens in his 81,274 syllables, 19 clear and 2 malformed, and none in the ministry's order. The pattern is rare; it turns up in 16 of the ministry dictionary's 50,105 syllables. Embree also marks the a in all 47 of his auh syllables, as in ka̍uh.

Until 14 October 2026 our Hokkien POJ and Tâi-lô converter read only the ministry's order and refused Embree's saⁿh as outside the ministry's tables. It put the tone 8 mark of auh on the u, and it respelled only the initial of a dialect final, which left output in neither system. The revision made that day reads both nasal orders, writes the ministry's in its output, marks auh on the a and keeps those dialect finals whole in Tâi-lô.

Taiwan's ministry hopes classrooms will employ Tâi-lô

The ministry announced the scheme on 14 October 2006 (民國95年10月14日), shortened to 臺羅. Its name for the language is 臺灣閩南語, Taiwan Southern Min. The minister's preface hopes, 希冀, that teaching and textbooks will adopt the scheme. Of other systems in private publishing and scholarship it says the ministry respects their existence, 尊重其存在. The manual's comparison tables head the POJ column 教會羅馬字, "church romanisation".

Advocates take sides. The ChhoeTaigi project's README calls POJ 台語ê正寫法, "the correct way to write Taiwanese", which is their claim.

Penang's Hokkien association works from a modified Tâi-lô

Persatuan Bahasa Hokkien Pulau Pinang describes its system, on a page dated 16 July 2021, as built from the Taiwan ministry's scheme and "appropriately modified to record local pronunciation of Penang Hokkien". Its site in October 2026 wrote Tsia̍h Î-pó and Kuí-kòo-sū, with Tâi-lô's ts and oo.

Penang has another system too. A page on penang-traveltips.com describes Taiji Romanisation, which swaps accents for tone numbers so it can be typed on any keyboard. It counts four tones plus a tone 33 that sounds like tone 3, gives checked syllables no tones of their own, and ends every syllable with its digit, as in kau4. For Singapore we found no body that sets a standard romanisation.

The ministry's 24 dialect finals stay off its Pe̍h-ōe-jī roster

Beside its 75 main finals, the manual lists 24 special ones for local accents and sound words, among them ir, ee, uinn and uang. It gives each a Tâi-lô spelling and no POJ column. In the dictionary's first readings they are scarce: 14 syllables of 50,105.

POJ writers did spell some of them. Embree writes the final uih 89 times, as in huih, and the manual lists uih among its special finals. The list also holds eng, illustrated with 兵 ("soldier"), and that one is a trap. Respell only its initial and Tâi-lô tseng becomes cheng, which is how POJ writes Tâi-lô tsing. For that syllable the ministry's table offers no POJ form to check against.

Advertisement
Eva Chin
Business & Chinese Culture Correspondent

Eva Chin covers business and commerce in Southeast Asia for RECATOOLS, alongside Chinese cultural practice and education.

View author profile → · Editorial policy

Corrections policy

Advertisement