Mova0.6 pre-release

the flag of South Korea 13. Numerals, quantifiers and classifiers

Middle Korean counts with native Korean and Sino-Korean numerals. A numeral can count a noun directly or state how many units of a classifier or measure occur. The same word can name an entity in one passage and supply the unit of counting in another. Numeral origin, noun meaning, and written form therefore need to be considered separately.

The Hangul examples below omit tone dots and use spacing to show the relevant grouping. Romanization represents the spelling, with ʌ for ㆍ and z for ㅿ. For characters printed without a Hangul reading, the romanization gives a Korean reading aid. The character alone does not establish its pronunciation in a particular passage.

Native Korean and Sino-Korean numeral series

The native series supplies ordinary counting forms. Its units combine with its tens in descending order. The 1459 vernacular preface counts the newly made letters with native twenty plus eight, followed by the character 字 ‘written character’. The spelling puts the two numerals together without a conjunction. 字 supplies the item counted (Hunminjeongeum eonhae 1459, § Vernacular preface).

FormRomanizationGlossTranslation
“스믈여듧 字”seumeulyeodeulp jatwenty-eight charactertwenty-eight letters

The table gives native counting forms for one through nine and for the multiples of ten through ninety. The forms used directly before a noun differ for some numerals, especially at the low end of the series. The counting forms in the table should therefore be read together with the forms in the section on order.

ValueUnitsRomanizationValueTensRomanization
1ᄒᆞ나hʌna10열yeol
2둘dul20스믈seumeul
3세sey30셜흔syeolheun
4네ney40마ᅀᆞᆫmazʌn
5다ᄉᆞᆺdasʌs50쉰swin
6여슷yeoseus60여슌yeosyun
7닐굽nilgup70닐흔nilheun
8여듧yeodeulp80여든yeodeun
9아홉ahop90아ᄒᆞᆫahʌn

Native vocabulary also includes 온 ‘hundred’ and 즈믄 ‘thousand’. These are separate numeral words, like the native tens. Their presence in the lexicon does not require a writer to use native forms for high counts. Written totals and formal enumerations can use the Sino-Korean series.

FormRomanizationGlossTranslation
온onhundredone hundred
즈믄jeumeunthousandone thousand

Sino-Korean numerals have a different internal structure. A unit before a scale word multiplies that scale, and a smaller expression following a larger one adds to it. In the ordinary formation of tens, 十 by itself gives one ten, and a preceding numeral specifies a higher multiple. Thus 十一 is ten plus one, 二十 is two tens, and 二十八 is two tens plus eight. Unfilled places can simply be omitted in a written cardinal. The same principles account for a written expression such as 三百五, with a hundreds term and a units term. A numerical expression in characters can be read as such in the text even when the witness gives no Hangul spelling of its reading.

ValueDigitKorean reading aidValueScaleKorean reading aid
1一il10十sip
2二i100百baek
3三sam1,000千cheon
4四sa10,000萬man
5五o
6六yuk
7七chil
8八pal
9九gu

In the introductory letter table of the Hunmong jahoe, the headings contain 八字 ‘eight letters’ and 十一字 ‘eleven letters’. The headings organize the letters by where they can be used in the syllable. The first count occurs for both the letters used in initial and final position and the letters used only initially. The second gives the count of medial letters (Choe Sejin 1527, § Eonmun jamo).

FormRomanizationGlossTranslation
“八字”pal jaeight charactereight letters
“十一字”sip-il jaten-one charactereleven letters

These witnesses also show how to identify the series in mixed writing. 字 follows a native numeral in the vernacular preface and Sinitic numerals in the letter-table headings. The origin of 字 does not determine which numeral series precedes it. A character heading also provides evidence for the written construction without settling how it would have been spoken in vernacular Korean.

Cardinals, ordinals and approximate quantities

A cardinal answers a question about amount. It can stand next to a counted noun, combine with a counting word, or stand on its own when the counted kind is recoverable. A numeral used on its own can also refer to the number itself. The context distinguishes naming the number two from saying that two members of a known group are involved. An overt noun or unit makes the domain of the count explicit.

An ordinal identifies position in an ordered set. Native 첫 ‘first’ modifies the following noun. In formal writing, 第 before a Sino-Korean cardinal marks an ordinal. The cardinal after 第 can itself be complex. In a written label, the name of the unit can precede the ordinal, as in a volume designation. The position of the label and the presence of 第 matter when the same characters could otherwise form a count.

ConstructionFormRomanizationGlossTranslation
Cardinal with a counted noun (constructed)두 ᄃᆞᆯdu dʌltwo monthtwo months
Native first position (constructed)첫 ᄃᆞᆯcheot dʌlfirst monththe first month
Written ordinal (constructed)第二je-iordinal-twosecond
Written label (constructed)卷第二gwon je-ivolume ordinal-twovolume two

Calendar labels can specify position by convention without 第. For example, 三月 in a date names the third month, whereas a quantity expression with the same characters can measure a duration of three months. The surrounding date formula or the clause in which the expression occurs establishes the reading.

An approximate count can give adjacent alternatives, a rounded estimate, or an explicitly open quantity. A close pair such as ‘two or three’ offers nearby alternatives and can be used for a rough estimate. It does not impose two successive counts on the noun. In written Sinitic expressions, 餘 after a number marks a remainder beyond that number. 數 in expressions such as 數十 gives an unspecified number of tens. These expressions leave different information open. A count of 十餘 is above ten, while 數十 gives the scale without an exact multiple. A bare cardinal can also be used loosely in context, although its form supplies no approximation marker.

ConstructionFormRomanizationGlossTranslation
Adjacent cardinals (constructed)두세 ᄃᆞᆯdu-sey dʌltwo-three monthtwo or three months
Cardinal with a remainder (constructed)十餘日sip-yeo ilten-excess daymore than ten days
Unspecified multiple of a scale (constructed)數十日su-sip ilseveral-ten dayseveral tens of days

Other quantifiers do not select a point on the numeral scale. 여러 ‘several’ leaves the count open, and 몃 ‘how many’ makes the count the questioned information. 몃 can also give an unspecified small number where the larger construction calls for an indefinite reading. 마다 following a noun distributes the statement over individual members of the relevant set. The 1459 preface uses it with ‘person’ in describing access to the letters. The phrase applies to each person without stating how many there are (Hunminjeongeum eonhae 1459, § Vernacular preface).

FormRomanizationGlossTranslation
여러 사ᄅᆞᆷ (constructed)yeoreo sarʌmseveral personseveral people
몃 사ᄅᆞᆷ (constructed)myeos sarʌmhow.many personhow many people
“사ᄅᆞᆷ마다”sarʌm-madaperson-eacheach person

The quantifier contributes a count, range, or distribution. Reference can come from a different part of the phrase. In 이 두 사ᄅᆞᆷ, the demonstrative selects an identifiable pair and the numeral gives its size. Without the demonstrative, 두 사ᄅᆞᆷ can introduce two people or refer to a pair already identifiable from the passage. Middle Korean has no obligatory article before a quantity phrase, and the numeral ‘one’ retains a counting use even when an English translation with a is natural. Interrogative quantity likewise differs from interrogative identity. Asking how many members are present leaves the size of the set open, while asking who is present leaves the identity open.

FormRomanizationGlossTranslation
이 두 사ᄅᆞᆷ (constructed)i du sarʌmthis two personthese two people
ᄒᆞᆫ 사ᄅᆞᆷ (constructed)hʌn sarʌmone personone person, a person

Counting words and their lexical restrictions

A numeral can count an entity named directly by a noun. In the preface's 스믈여듧 字, 字 identifies the entities being counted. Where the surrounding discussion supplies a general kind, a more specific counting word can tell the reader what counts as one item. A counting word can also select a conventional unit of a substance, an interval, or an object made up of smaller parts. These uses overlap with ordinary nouns. A word that names a volume can count volumes of a work, and a word that names a day can specify the unit of a duration.

The lexical restriction comes from the unit. 字 counts characters in the preface and in the letter headings. The count is of letter types there because the passages specify an inventory. In another context a count with 字 can concern written occurrences. 點 counts marks in the preface's directions on tone. One and two marks distinguish the two marked categories described there, and the directions also distinguish the case without a mark. These are counts of inscribed marks rather than of letters or syllables (Hunminjeongeum eonhae 1459, § Tone-mark directions; Choe Sejin 1527, § Eonmun jamo).

FormRomanizationGlossTranslation
“ᄒᆞᆫ 點”hʌn jeomone dotone dot
“두 點”du jeomtwo dottwo dots

In a phrase with both an entity noun and a counting word, their meanings jointly determine the count. With a horse noun, 匹 counts individual animals in the appropriate class. With a book or work, 卷 counts volumes or comparable textual divisions. A count in 卷 can differ from a count of complete works, even when the same set of physical objects supplies both totals. A quantity measured in ᄃᆞᆯ counts months of an interval. The counted interval need not contain multiple objects.

ConstructionFormRomanizationGlossTranslation
Animal counted with a sortal (constructed)ᄆᆞᆯ 두 匹mʌl du pilhorse two headtwo horses
Work counted by volumes (constructed)冊 두 卷chaek du gwonbook two volumetwo volumes of a book
Duration measured in months (constructed)세 ᄃᆞᆯsey dʌlthree monththree months

The interpretation of a counting word can survive omission of the entity noun. In a passage about horses, a count of two with 匹 can draw its referent from the preceding text. A volume count likewise needs no repeated name for the work when the context supplies it. Conversely, an overt entity noun can supply the countable unit by itself. People can be counted with 사ᄅᆞᆷ ‘person’, and the preface counts letters with 字. The inventory headings in the Hunmong jahoe use 字 directly after the numeral and leave the kind of letter to the rest of the heading (Hunminjeongeum eonhae 1459, § Vernacular preface; Choe Sejin 1527, § Eonmun jamo).

These lexical restrictions do not create a system of grammatical gender. The counting word can be sensitive to the kind of entity or the way it is measured, while the other words in the phrase show no gender agreement. A noun following a numeral greater than one can also keep the same form it has after ‘one’. The numeral already supplies the relevant cardinality. Where a plural or collective expression occurs, its grouping or discourse contribution must be interpreted separately from the numeral.

Order of noun, numeral and classifier

With direct counting, the numeral can precede the noun. This is the order in the vernacular preface's native count of letters. The same linear order, numeral followed by counted unit, appears inside the Sinitic counts in the letter-table headings. With an overt entity noun and a distinct counting word, a count can follow the entity noun, with the numeral before its counting word. When the noun is understood, the numeral and counting word can make up the overt phrase. An entity noun can also precede a numeral used without a separate counting word. In that order, the numeral can present a tally of the entities named by the noun (Hunminjeongeum eonhae 1459, § Vernacular preface; Choe Sejin 1527, § Eonmun jamo).

PatternFormRomanizationGlossTranslation
Numeral + counted noun (constructed)두 사ᄅᆞᆷdu sarʌmtwo persontwo people
Entity noun + numeral (constructed)사ᄅᆞᆷ 둘sarʌm dulperson twotwo people
Entity noun + numeral + counting word (constructed)ᄆᆞᆯ 두 匹mʌl du pilhorse two headtwo horses
Numeral + counting word, entity understood (constructed)두 匹du piltwo headtwo horses, in a context about horses
Numeral, counted kind understood (constructed)둘dultwotwo of them

The position of a low native numeral can affect its form. ᄒᆞ나 and 둘 are counting forms that can stand on their own. ᄒᆞᆫ and 두 are their forms before the noun or counting word in the constructions illustrated here. 세 and 네 can occur before nouns, while the restricted forms 석 and 넉 occur before particular units, including the month word ᄃᆞᆯ. These restrictions belong to individual numeral and unit combinations. They cannot be predicted by treating every noun beginning with the same sound alike. The independent spellings below omit inflection, where the stem can show additional material before a particle.

ValueCounting formRomanizationPrenominal formRomanizationRestricted prenominal formRomanization
1ᄒᆞ나hʌnaᄒᆞᆫhʌn
2둘dul두du
3세sey세sey석seok
4네ney네ney넉neok
FormRomanizationGlossTranslation
석 ᄃᆞᆯ (constructed)seok dʌlthree monththree months
넉 ᄃᆞᆯ (constructed)neok dʌlfour monthfour months

Other modifiers can narrow the set over which the numeral counts. A demonstrative before the numeral can identify the group. A possessor before a numeral and noun can establish whose entities are being counted. These modifiers contribute reference to the whole count. The numeral supplies the cardinality, and the counted noun supplies the domain. A count can also be stated separately for an already established set. In running text, particles and the surrounding clause help decide whether a sequence of noun and numeral forms one quantity phrase or whether the numeral is a separate tally.

FormRomanizationGlossTranslation
그 두 사ᄅᆞᆷ (constructed)geu du sarʌmthat two personthose two people
내 두 아ᄃᆞᆯ (constructed)nay du adʌlI.GEN two sonmy two sons

The spaces in these examples show the grammatical grouping. The source form 스믈여듧 itself shows that a boundary between the tens and the units need not appear as a space. Likewise, a mixed-script boundary between a numeral and a unit does not by itself divide them into separate phrases.

Case and discourse particles with quantity phrases

A quantity phrase takes the case appropriate to its role in the clause. In a direct count with the numeral before the noun, a case particle attaches to the noun. In a unit construction whose phrase ends in a counting word, the particle can attach to that unit. In the vernacular preface, the object ending follows 字 at the outer edge of the count ‘twenty-eight letters’. The numeral and the unit together fall within the scope of the case particle (Hunminjeongeum eonhae 1459, § Vernacular preface).

FormRomanizationGlossTranslation
“스믈여듧 字ᄅᆞᆯ”seumeulyeodeulp ja-rʌltwenty-eight character-ACCtwenty-eight letters, as object
두 사ᄅᆞ미 (constructed)du sarʌm-itwo person-NOMtwo people, as subject
두 사ᄅᆞᄆᆞᆯ (constructed)du sarʌm-ʌltwo person-ACCtwo people, as object
두 卷을 (constructed)du gwon-eultwo volume-ACCtwo volumes, as object

The noun or unit to which the particle attaches determines its form under the ordinary rules of nominal inflection. In 사ᄅᆞ미 and 사ᄅᆞᄆᆞᆯ, the stem-final consonant is written in the following syllable with the vowel of the ending. The analysis in the gloss keeps the consonant with the stem. A numeral used independently can take a particle itself. If the relevant numeral stem has a final ㅎ that is visible before an ending, the independent inflected form can show it. The form 둘히 illustrates this with nominative 이.

FormRomanizationGlossTranslation
둘히 (constructed)dulh-itwo-NOMtwo of them, as subject
두 사ᄅᆞᄆᆞᆫ (constructed)du sarʌm-ʌntwo person-TOPas for the two people
두 사ᄅᆞᆷ도 (constructed)du sarʌm-dotwo person-ADDtwo people too, even two people

Topic and additive particles can have scope over the whole count. A topic-marked phrase may establish which counted group the statement concerns or contrast its quantity with another. 도 can include the counted group alongside another group, or give an ‘even’ reading when the context ranks the quantities or groups as more or less expected. Its presence alone does not settle which reading is intended. Similarly, an unmarked numeral gives no grammatical indication of whether the numeral itself bears contrastive focus. The discourse can contrast two people with three people, or contrast one pair with a different group of the same size.

A separate tally can follow a noun phrase that already carries a particle. In that configuration the particle marks the preceding phrase. The following numeral still has to be interpreted in relation to the predicate and the entities being counted. When the sole particle follows the final counting word, the placement instead supports treating the noun and the unit count together as the phrase in that grammatical role. A case-marked duration also calls for attention to the predicate. An accusative-marked expression meaning ‘three months’ can measure how long an event lasts, while an accusative-marked object count specifies the quantity of the affected entities.

Possession and quantity can occupy different levels in the same phrase. A possessor inside the phrase narrows the counted set, as in 내 두 아ᄃᆞᆯ. The entire quantity phrase can in turn serve as a possessor, in which case the genitive belongs at its outer edge. Neither a genitive nor a topic particle supplies a number value. A demonstrative, a possessor, or preceding discourse can establish the referent even when the count itself is the only overt description later in the passage.

Measures, dates and textual conventions for counting

A measure gives the amount of something in a named unit. Time words can measure duration, count separate intervals, or label a calendar position. The noun determines the unit, while the construction and passage determine how to interpret the numeral. With native counting, 두 ᄃᆞᆯ naturally expresses two months of duration in an appropriate clause. In a written calendar formula, 月 after a numeral marks the month field and 日 after a numeral marks the day field. Translate these as calendar positions when the formula locates a date. A conversion to a Gregorian date requires the calendar and the relevant year, including identification of any intercalary month.

UseFormRomanizationGlossTranslation
Duration (constructed)두 ᄃᆞᆯdu dʌltwo monthtwo months
Month and day in a date (constructed)三月十五日sam-wol sibo-ilthree-month ten-five-daythe fifteenth day of the third month
Intercalary month and day in a date (constructed)閏三月一日yun sam-wol il-ilintercalary three-month one-daythe first day of the intercalary third month
Duration in a written expression (constructed)一日il-ilone-dayone day
Day field in a date (constructed)一日il-ilone-daythe first day

An era or reign designation can provide the starting point for a numbered year, with 年 identifying the field. A sexagenary year name can also locate a year within a cycle. A reader needs the chronological setting to identify the particular cycle intended. In a date with both a year designation and the month and day, each number belongs to its own field. The sequence represents one date. In a duration, the unit words instead contribute quantities that may be combined to express the length of an interval.

Counting in a book can concern several different objects. A volume number locates a division in a sequence. A total in volumes measures the extent of a work or the size of a collection. A count of letters in a table can concern either the entire inventory or a class named in a heading. In the Hunmong jahoe, the two headings with 八字 count different sets, as shown by the distinction between shared initial and final use and initial-only use. Its 十一字 heading counts medial letters. The 1459 vernacular preface gives an overall total for the newly made letters. The scope of each count comes from what the passage or heading is counting (Hunminjeongeum eonhae 1459, § Vernacular preface; Choe Sejin 1527, § Eonmun jamo).

Hangul and characters can occur together within one quantity phrase. In the preface, the numeral is in Hangul, the unit is a character, and the following case marker is in Hangul. The Hunmong jahoe headings write both numeral and unit in characters. A transcription can make the grouping legible with spacing and still retain those script choices. When a numeral and its scale are written as a single run, the arithmetic follows the structure of the numeral series. Joining the written forms of ten and one in 十一 gives eleven, while joining two and ten in 二十 gives twenty. Under the native series, joining twenty and eight in 스믈여듧 gives twenty-eight (Hunminjeongeum eonhae 1459, § Vernacular preface; Choe Sejin 1527, § Eonmun jamo).

The actual marks on a page also matter. In the preface's directions for tone, 點 names a graphical dot that can itself be counted. A tone dot placed beside a Hangul syllable on the page belongs to the notation of that syllable. In a transcription, a mark reproduced beside a spelling and the word for the mark in an instruction therefore represent different kinds of information. The same principle applies to a volume number supplied in a heading and a volume number supplied by a modern edition. Only the former provides evidence for how the witness wrote its label (Hunminjeongeum eonhae 1459, § Tone-mark directions).

How Mova handles numerals and classifier phrases

Middle Korean (okm) has no level-2 parser model. In multilingual parsing, tools/langseg assigns the language of sentences and spans at level 1. Hangul and Han characters are outside its supported script and lexicon categories and are assigned other. At level 2, tools/mova-parse reports the missing model in the # parse comment. Its output for okm leaves LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL as _ and sets Parse=None in MISC.

Mova does not tokenize, tag, lemmatize or parse Middle Korean. The numeral forms, classifier restrictions, phrase orders, and uses of particles described here have no implemented okm analysis. Mova has no annotated Middle Korean treebank and no measured treebank baseline. Its frame, pragmatic and world components read English trees and have no Middle Korean parse to read.

Squid reads the chapter's grammatical points as claims and queues formalizable claims for possible implementation. Queued claims are pending. The first points for implementation are those needed by a tokenizer, a tagger and a Universal Dependencies treebank baseline.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/ko/okm/13-numerals-and-classifiers/