Wikipedia:WikiProject Korea/Romanization of Korean on Wikipedia

Romanization of Korean on Wikipedia

Since the first version of the MOS in 2004, there has been regular confusion and debate over how to handle all of these issues on Wikipedia. These discussions have often been repetitive and wasteful.

In 2024, WikiProject Korea wrote a complete rewrite of the MOS. This effort and its relevant research and argumentation is documented in this essay.

Why not use one romanization system for everything?

While tempting and this would be easier, this would go too far against our § Basic principles.

Why not use other systems?

We try to balance reflecting common practice and reducing complication. The systems we currently recommend are more than sufficient for most cases; asking users to learn more systems adds significant complication without adding much understanding.

Why use MR for pre-1945 topics?

International academic journals almost universally use MR when writing about Korean history.^[6][7] Practice in books seems to be more divided but still leaning MR, per § Romanizations used in books.

Why use 1945 as a dividing line for using RR?

Mid/late August 1945 was when the division of Korea occurred.

Other possible lines:

• 1948 (the establishment of North/South Korea)
  • Arguably less significant than the 1945 division. Little changed between 1947 and 1949, but 1944 to 1946 was a seismic change.
  • Similarly, dividing spellings for terms relating to the United States Army Military Government in Korea (1945–1948) and South Korea (1948–) seems relatively arbitrary. There was relatively clear continuity between both these entities.
• 2000 (the enactment of RR)
  • Without a doubt, less significant than the division of Korea. Notability is unlikely to change from 1999 to 2001.
  • Some recent writings use MR when talking about post-division South Korea.

Why use RR for South Korea?

RR is now widely used by the international and domestic English-language press for coverage of South Korea. South Korea's popular culture and soft power have boosted the use of the system. Books on contemporary South Korea also often employ the system.^[d]

Why use Yale romanization?

The international linguistics community still seems to widely use Yale.^[8] Restricting its use would mean asking our editors to convert Yale text in sources to either MR or RR; this is difficult to enforce, leaves room for errors, would be misrepresenting current practice, and adds complication.

How extensively or firmly Yale should be recommended is currently uncertain.

Why not use North Korea's romanization for North Korean articles?

North Korean romanization ("NKR") is used by the NK government and state media. It is also recommended by the AP Stylebook for people names and by NK News for general cases (see § Other style guidelines).

Arguments against using NKR:

• The international academic community widely uses standard MR when writing on North Korea.^{[verification needed]}
• AP and NK News aren't consistent in their usage of NKR.
• It is based on MR,^[7] which it is also reasonably similar to.
• Minimizing the number of systems in use on Wikipedia.
  • This is currently weighted highly.
• Using standard MR has been the status quo since the first version of the MOS in 2004, and seemingly nobody has disputed it until 2024.

Why use the 1961 version of MR?

This version appears to be the most widely used.^{[verification needed]} The 1961 paper is effectively a summary of the 1939 version, with minor updates to reflect more recent Korean phonology and spelling. It is missing some details that the original had (as it is a summary). We assume those details still apply; they have not been affected by the development of the Korean language.

How strictly the specifications of MR are followed seems to vary; some authors claim to use the system but romanize in ways that are either incorrect or recommended against in the original 1939 and 1961 papers.

Why not use the ALA-LC (Library of Congress) version?

This version appears to not have widespread adoption in academic works. It's also extremely complicated, with countless instructions for edgecases.

The ALA-LC version heavily focuses on word division in phrases and sentences. While that may be important when romanizing book titles in libraries (and for easier word-based searching in library databases), that is much less important on Wikipedia, where phrases and sentences are usually translated into English (though romanizations can also be given).

In addition, the ALA-LC version has strange rules about two-syllable given names: non-Sino-Korean names have joined syllables, and Sino-Korean names have a hyphen between the two syllables. This is needlessly complicated and not really possible to follow. It is not really possible to determine which names are Sino-Korean and which are not. Even ALA-LC admitted this.^[9] The original MR does not do this – it treats given names just like any other words.

Hyphenation

The 1939 paper that introduced MR says the following about hyphenation:

in Korean, a given syllable may vary widely in pronunciation according to the nature of the other syllables in the word [...]. A simple example, the word Silla, will help to clarify the point. In Chinese, hsin 新 plus lo 羅 are pronounced Hsin-lo but in Korea, sin 新 plus na (la) 羅 are pronounced Silla. To hyphenate this name as Sil-la would imply that it is composed of two parts which individually are sil and la, which is obviously misleading.^[10]

There are many cases where there is a partial division does not seem great enough to justify writing them as separate words. It is for such cases that we have reserved the use of the hyphen; but we believe that on the whole it should be used as sparingly as possible. [Example: 연산군 → Yŏnsan-gun]^[11]

The 1961 summary paper on MR says the following about hyphenation:

The hyphen is reserved for subsidiary divisions within words, such as Tŏksu-gung (덕수궁), Haein-sa (해인사).

Reischauer, one of the namesakes of MR, expressed general distaste for the use hyphens for pronunciation-related matters. When the Japanese government introduced a Japanese romanization rule (1937) that employs a hyphen when the moraic n (ん) is followed by a vowel (including y), he wrote the following (excerpt from Rōmaji or Rōmazi (1940)):

Another change for the worse in Kokutei is the substitution of the hyphen for ' between the final n of one syllable and an initial vowel in the next syllable within a single word. The hyphen is best reserved for use in words which should be hyphenated on other than phonetic grounds.

It could be argued that the above is part of the reason why MR uses n'g (not n-g) when n and g are pronounced separately (e.g. 한글 Han'gŭl), despite the apostrophe also being used for strongly aspirated consonants. "Mun'gyŏng-si" is much easier to parse than "Mun-gyŏng-si"; the hyphen is more suitable for semantic division.

Spacing

In MR, there are mandatory and optional spaces.

Mandatory spaces are just those between the family names and given names of people.^[12]

Optional spaces are more subjective. The 1939 paper that introduced MR says the following:

In Romanizing sentences, book titles and the like, the problem of a correct division into words is important. It is difficult to define what is a word in any language, and the problem is particularly complicated in Korean where divisions have been traditionally made by syllables and by clauses. Therefore in Romanizing groups of words a division into units roughly comparable to those in European sentences should be made if the Romanization is to be intelligible to the average Occidental.

One can not lay down hard and fast rules for word divisions, and much depends on the special considerations and needs of each individual problem of Romanization.^[13]

The paper then gives some recommendations on space insertion:^[14]

• For native Korean terms, no space between verb stems and verb endings (e.g. 자다 ("to sleep") → chada and not cha ta)
• For native Korean terms, no space between nouns and postpositions (e.g. 꽃이 ("flower" + subject postposition) → kkoch'i and not kkot i)
• For other remaining native Korean terms (e.g. adjectives, adverbs, etc), attempt to follow spacing practices of English.
• For Sino-Korean words, recommendations are difficult; rely on experience, typical groupings of characters, and to maximize understanding to readers.

They provide this example:^[15]

한글 운동은 연산군 조에 이르러 큰 액운을 당하였다.
Han'gŭl undongŭn Yŏnsan-gun choë irŭrŏ k'ŭn aegunŭl tanghayŏtta.
(The Han'gŭl movement was placed in an extremely critical position during the reign of the ruler Yŏnsan.)

Common MR mistakes

• 시 should be romanized ⟨si⟩ (not ⟨shi⟩), per the original MR. ⟨sh⟩ only appears when romanizing 쉬 (⟨shwi⟩).^[16]
• Syllable-final ㄹ + syllable-initial ㅎ is ⟨rh⟩ (not ⟨lh⟩).
• The apostrophe only appears in ⟨ch'⟩, ⟨k'⟩, ⟨p'⟩, ⟨t'⟩, and ⟨n'g⟩. It does not appear elsewhere.
  • When ⟨n⟩ and ⟨g⟩ are pronounced separately, an apostrophe is added between them (i.e. ⟨n'g⟩); ⟨ng⟩ is not followed by an apostrophe.

Hyphenation

In RR, hyphens can be either optional or mandatory.

• Optional hyphens are used in two scenarios:
  1. Disambiguating pronunciation (e.g. 해운대 Hae-undae)
  2. Between syllables of a given name (e.g. 홍길동 Hong Gil-dong)
• Mandatory hyphens are for separating an administrative unit (e.g. 평창군 Pyeongchang-gun)

In general, we recommend against the use of hyphens for disambiguating pronunciation. This seems to align with the explanations from the National Institute of Korean Language.^[5][17][18]

Also, mixed use of hyphens can be confusing. For example, in Hae-undae-gu, the first hyphen is for disambiguation in pronunciation, but the second one is for separating an administrative unit. Consistently excluding optional hyphens (i.e. Haeundae-gu) makes names more immediately recognizable.

Note that hyphenation in the names of geographic features, cultural properties, and man-made structures is prohibited.^[19]

For hyphenation in people names, see § RR names.

Spacing

RR requires a space between family and given names. It does not officially comment on optional spaces.

Letter-by-letter transliteration

When reversibility is desired (namely in academic articles), RR officially allows for a letter-by-letter transcription. For example, 독립 would be rendered as "doglib" in the letter-by-letter transcription, whereas by its normal pronunciation spelling it'd be "dongnip".^[19]

In general, we discourage the use of such transcriptions. Arguments against using it:

• They are an alternate form of RR and less commonly used (thus making them less useful and recognizable to the majority of people).
• In any situation where they may have some benefit, the original Hangul could be added instead. Hangul is more recognizable than letter-by-letter RR.
• While it is tempting to try and advocate for our consistent use of this system because of its reversibility with Hangul and ease of use for people who know Hangul, remember our § Basic principles.

Why have these steps for WP:KOREANNAME?

Over the past 20 years on Wikipedia, there has been constant costly debate and confusion over how to weigh WP:COMMONNAME, personal preference of the subject, and systematic romanization for Korean names. These debates have often reached differing conclusions, and have created a confusing mess of inconsistent practice. We established these steps to hopefully reduce these debates and improve WP:TITLECON.

Note that this process was designed to consider all five WP:CRITERIA for titles:

• Recognizability – accounted for in all three steps. In the third step, we even chose to use a common modification of RR (hyphens in given names) to improve recognizability. WP:KOREANNAME is designed to match most common practice for various common subject matters and time periods.
• Naturalness – accounted for in all three steps. Same explanation for above.
• Precision – self-explanatory
• Concision – self-explanatory
• Consistency – accounted for mostly in the third step. First two steps yield exceptions, the third step defaults to a safe consistent practice.

Why ask for "widely used" WP:COMMONNAMEs in step 1?

Multiple reasons.

1. For cases when there are extremely few attestations in English, there are so many conflicting romanization practices that taking a small sample will likely yield inconsistent romanization (including possibly ad-hoc romanization), which hurts WP:TITLECON and potentially WP:RECOGNIZABILITY. We'd rather people turn to a default spelling to avoid those downsides.
2. When there is significantly divided use, while we'd love for if people really researched each concept to determine which names are 51% vs 49% more common, this is costly and people rarely want to engage in such research. So we give them an easy out by turning to a default practice.

MR names

The 1939 MR paper says this about people names:^[12]

In both two character surnames and two character given names the general rules of euphonic changes should be observed, and the two syllables should be written together... [For assimilating between surnames and given names,] disregard euphonic changes between surnames and given names or titles... [For the surname 李], it is pronounced in the standard dialect and should be Romanized I, but some may prefer to retain the older Romanization, Yi, because that is already the familiar form.

We recommend following the above guidance; it also appears to be commonly followed in academia. Also, "Yi" is more likely to be the WP:COMMONNAME spelling, and is also arguably more readable than "I".

For hyphenation in names, we currently do not recommend in WP:KOREANNAME that hyphens be inserted. This is subject to change; it is not yet clear to us whether hyphenation or non-hyphenation of MR names is more dominant (some research documented in #Romanizations used in books).

RR names

Assimilated sound changes are not reflected in given names.^[19]

For given names, hyphens are officially discouraged by the National Institute of Korean Language^[19] and the Ministry of Culture, Sports and Tourism.^[20] See also Revised Romanization of Korean § Hyphenation. However, English-language press in both South Korea and abroad appear to widely use them. Per our § Basic principles, we choose to hyphenate.^[e]

On the official RR documentation, it says "Transcriptions of family names will be established additionally.", implying that they are planning to eventually establish additional guidelines on how to romanize family names. However, even into 2026, this hasn't come to fruition.^[19] Since it has been already 25 years, this may actually never happen.

Common English spellings for South Korean surnames

For South Korean surnames in English-language text, if there's evidence of a widespread common spelling we recommend the use of that spelling.

We choose to do this instead of the alternative: defaulting to standard RR spellings if common name or personal preference are not known. Defaulting to RR may have some benefit in recognizability and consistency. However:

• For the surnames where almost all people use a non-RR spelling (e.g. Park for 박; Bak), we'd be wrong almost all of the time.
  • Per naturalness, people expect to see these spellings for those names, even if they're ad hoc.
• Few people seem to actually know how RR works in detail, meaning recognizability may be weak.

How widespread a surname spelling should be before we recommend it is not certain. We also only included the top N most common surnames in order to not make the table overly lengthy. Initial commonality threshold is 80%; no prejudice against lowering it or increasing N.

North Korean romanization

At present, we do not use North Korean romanization on Wikipedia except for cases of WP:COMMONNAME. We considered applying the rules of North Korean romanization for names to MR, but it was argued that doing so was too arbitrary and inconsistent.

Regardless, below is what North Korean romanization recommends for people names. We document this here for ease of access.

Spaces are given between the first and second names of a person and the first sounds of the names are written in capital letters and in the names of Chinese character origin, spaces are given between each letter. When the first sound of the second syllable becomes vocal, it is transcribed in a relevant voiceless letter.

e.g.

• 김꽃분이 Kim KKotpuni
• 박동구 Pak Tong Gu
• 안복철 An Pok Chŏl^[24]

Partial tautological names

One point of uncertainty is whether to use terms like "Gyeongbokgung Palace" ("gung" means "palace") or "Bulguksa Temple" ("sa" means "temple"). Beginning on January 1, 2013, the South Korean government body Cultural Heritage Administration began officially preferring that style.^[25] The Seoul Metropolitan Government also shares this preference. However, this practice has varying levels of adoption elsewhere and in various newspapers. For example, the Korea JoongAng Daily has explicitly rejected it.^[26]

In general, our naming conventions currently discourage the practice. Various Wikipedians have expressed distaste for it, as it seems repetitive if one speaks Korean and English. However, it has also been pointed out that such repetitions are not "incorrect" in linguistics, and are actually quite common elsewhere in the world.