Erika RIMKUTĖ

Vytauto Didžiojo universitetas

HOMONYMS, HOMOFORMS AND HOMOGRAPHS FROM THE PERSPECTIVE OF CORPUS LINGUISTICS

1. Introduction

Vytautas Magnus University's Centre of Computational Linguistics has compiled and continues to expand the Corpus of the Contemporary Lithuanian Language, which contains more than 100 million words (address: http://donelaitis.vdu.lt/Tekstynas). A morphologically annotated corpus of one million words is also being compiled and maintained. For this purpose, the computer program Lemuoklis, developed by V. Zinkevičius (Zinkevičius 2000: 246), is used. It can provide the headword form of a written word form, known as the lemma, and indicate morphological tags. For example, the form laužo is lemmatized as the noun laužas or the verb laužyti, and is identified as either the genitive singular of a masculine noun or the third-person singular present indicative of a verb. 47 percent of all words or word forms are morphologically ambiguous. Morphologically ambiguous words or word forms are those for which Lemuoklis automatically identifies two or more lemmas or provides two or more possible morphological tags, e.g.: namo – noun and adverb (nãmo stogas – einu namõ); galimas – adjective and participle (galimas daiktas); laimės – noun and verb (lái-mės kūdikis – jis laimės varžybas), and so on.

Texts that have been automatically lemmatized and morphologically tagged are then reviewed by a linguist, who eliminates ambiguous cases. Such ambiguous forms have been analyzed in greater detail. This is a new area of research, so the terminology needs to be selected and defined. In linguistics, these forms are referred to in various ways: grammatical / morphological homonyms, homofoms, homographs, partial homonyms. This article aims to reveal inconsistencies in the use of existing terms and to justify the use of the term homoforms and the new term morphological ambiguity.

33

2. Concepts of Homonymy

Until now, morphological ambiguity has received little attention in Lithuanian linguistics; lexical homonyms alone have been examined in detail. This has led to problems in naming morphological ambiguity and its individual types.

2.1. Homoforms and morphological homonyms. In linguistics, morphological / grammatical homonyms are described as phonetically identical forms of the same word that have different grammatical meanings, e.g.: ranka (singular nominative) and ranka (singular instrumental) (Jakaitienė 1980: 42; Соболева 1980: 78). Some linguists do not distinguish grammatical homonyms from homoforms. These may be different forms of the same word (see the examples mentioned above) or similarly sounding forms of different words, e.g.: sakai (noun, nominative) and sakai (verb, present tense, singular, second person) (LKK I 1995: 23; Barauskaitė 1979: 43–44; Drotvinas 1986: 125). F. Katamba calls identical forms of the same or different parts of speech syncretism (Katamba 1994: 25).

E. Jakaitienė homoformų netapatina su morfologiniais homonimais (Jakaitienė 1980: 42). Homoformas apibrėžia kaip atskiras, atsitiktinai sutampančias bendrinėje kalboje savo fonetine struktūra visiškai skirtingų žodžių formas, pvz.: kiškis (dkt. vns. vard.) ir kiškis (sngr. vksm. liep. nuos. vns. II asm.). Homoformomis laikytini ir tokie žodžiai, kurių fonetiškai sutampa ne viena, o kelios formos. Pavyzdžiui: a) veiksmažodžiai, kurių fonetiškai sutampa tik bendratis ir iš jos padarytos formos: gesti, gęsta, geso „ugniai, šviesai nykti, blėsti“ ir gesti, genda, gedo „darytis netinkamam“; b) daiktavardžiai ir būdvardžiai ar dalyviai (skiriasi jų vienaskaitos naudininko ir vietininko, daugiskaitos vardininko ir naudininko formos): bendras „priklausomas visiems“ ir bendras „bičiulis, draugas“ (Jakaitienė 1980: 43). LK I tokie atvejai priskirti prie dalinių homonimų, tiesa, ten minimi tik veiksmažodžiai (LK I 1995: 24).

LK I prie homoformų aptarti fonetiškai sutampantys linksniai (LK I 1995: 23), bet kyla klausimas, kuo laikytini sutampantys linksniai, kurie skiriasi tarimu, pvz.: naujos parduotuvės atidarymas – atidarytos dvi naujos parduotuvės. V. Drotvinas tų pačių kalbos dalių ir vienodai, ir skirtingai tariamas formas vadina reguliariosiomis homo-

34

formomis (Drotvinas 1986: 93). Kiti kalbininkai neužsimena apie to paties žodžio skirtingai tariamas formas.

2.2. Homoforms and partial homonyms.

E. Jakaitienė defines partial homonyms as words in which all the forms of one word coincide with some of the forms of another word. Most often, one word in such a pair of homonyms belongs to an inflecting part of speech, while the other belongs to an uninflected one, e.g.: greta—a noun and an adverb, deja—a noun and an interjection (Jakaitienė 1980: 43).

Iš LK I pateikto dalinių homonimų apibrėžimo (LK I 1995: 24) galima spręsti, kad ši homonimijos rūšis sutampa su anksčiau jau E. Jakaitienės minėtomis homoformomis (Jakaitienė 1980: 43), pvz.: rausti, rausia, rausė „kasti, knisti“ ir rausti, rausta, raudo „darytis raudonam“.

KTŽ classifies as homoforms the forms of different words that are pronounced and written identically (KTŽ 1990: 79). Thus, several coinciding case forms of the same word are not classified as homoforms. With homoforms defined in this way, the distinction between homoforms and partial homonyms becomes unclear, since the latter are described as “homonyms whose morphemes do not all coincide completely or which belong to different parts of speech” (KTŽ 1990: 80). This type of homonym is illustrated by the example greta—a noun and an adverb, while the example given for homoforms is varu—an adverb and the instrumental case of the noun varas (KTŽ 1990: 79–80); consequently, no distinction between partial homonyms and homoforms is apparent.

2.3. Homoforms and homographs.

The greatest uncertainty for the author of this article arose when analyzing forms of the same or different words that differ in prosodic features, i.e., stress placement, pitch accent, or voice quantity: what should they be called—homographs or homoforms? Homographs are most often defined as different words or their forms that are written identically but pronounced differently (differing either in pitch accent or in stress placement), e.g.: austi “to cool down, become less hot” and aūšti “to dawn, to grow light” (Jakaitienė 1980: 44; LK I 1995: 23; KTŽ 1990: 23–24; Palmer 1981: 101; Drotvinas 1986: 91–92; Barauskaitė 1979: 44). Thus, it would seem that ambiguous forms differing in prosodic features should be called homographs, but LK I states that “there are also homoforms that are written identically but differ in stress” (LK I

35

1995: 23). The following examples are given: “baltų and baltų (báltai and báltas), dúovanos and dovanõs (dovanà and dovanóti), māno and māno” (LK I 1995: 23). V. Drotvinas lists the examples supánčios (verb, future tense, third person) and supančios (participle, feminine, singular genitive or plural nominative) as homoforms (Drotvinas 1986: 91). These examples also differ in stress placement, although they are classified as homoforms.

The concept of homoforms differing in stress placement seems logical, since analysis of morphologically annotated texts has revealed many forms that differ in pronunciation (about 20 percent of homoforms differ in prosodic features) (see Rimkutė 2002: 91–92). But the question arises as to what criteria can be used to distinguish homoforms from homographs if phonetic identity is not required for homoforms. Perhaps only forms of the same word that differ in pronunciation should be considered homoforms, e.g.: mamà (singular nominative) and mãma (singular vocative), galì (verb galėti, present tense, second person singular) and gãli (verb galėti, present tense, third person singular), while homographs would be words with different meanings, e.g.: kártis “a pole” and kàrtis “bitterness,” or kártis “a horse’s neck; the place where a mane grows.” However, the examples given in LK I and by V. Drotvinas are not forms of the same word that differ in prosodic features, but words with different meanings. Moreover, it is not always clear whether these are forms of the same word or of different words.

Thus, it would seem that different words or word forms that differ in prosodic features may be considered homoforms. Another criterion that explains this conception of homoforms is that homographs are often described as words with different meanings, although not all linguists emphasize this; for example, LKE states that homographs are words or word forms written identically (LKE 1999: 248). Several case forms of the same nominal word often coincide. It goes without saying that these are not words with different meanings. It is therefore unclear which type of homonymy such forms should be assigned to.

O. S. Akhmanova calls cases in which forms of the same word are pronounced differently homographic word forms and emphasizes that these are not homographs (Ахманова 1986: 436). V. Drotvinas proposes calling homoforms, homographs, and homophones homonymous word forms (Drotvinas 1986: 91). This term seems acceptable, so those word forms that differ in prosodic features would

36

may be called homographic or homonymous word forms, but instead of the long term, a shorter one was chosen—simply homoforms. It has already been mentioned that the analysis is partly determined by the object of study—a morphologically annotated corpus annotated automatically—so phonetic differences between homoforms are not examined in depth, since this requires more extensive research. These criteria led to a slight modification in the use of the term homoforms.

To avoid confusion when analyzing morphologically ambiguous forms in a morphologically annotated corpus, the assumption is made that inflected forms assigned to the same or different parts of speech, which differ in stress placement, pitch accent, or vowel quantity, may be both homoforms and homographs, e.g.: the king’s throne – he ruled for a long time, the roof of the house – I am going home, the elders are sitting – it happened long ago. Thus, some homoforms—identical nominal case forms—may be homographs when considered from the perspective of pronunciation, while homographs may be homoforms or different case forms of the same words when examined morphologically. The simplest option was chosen: all morphologically ambiguous forms, whether or not they differ in prosodic features, are considered homoforms, and the term homographs is not used at all.

2.4. Other cases of homonymy.

In some linguistic studies, the scope of homoforms is not entirely clear: are homoforms made up only of word forms belonging to the same part of speech, or also of forms belonging to different parts of speech? Based on the examples provided by V. Urbutis and LKE, it can be inferred that some identical word forms belonging to different parts of speech are considered homoforms, e.g.: grūdo—the singular genitive of the noun grūdas, and grūdo—the third-person singular past frequentative form of the verb grūsti (Urbutis 1956: 4; LKE 1999: 247–248); it is therefore unclear whether identical word forms belonging to the same part of speech can be considered homoforms, for example, tapo (the other principal forms differ: tap yti, tapė and tapti, tampa).

The definition of homographs provided in LKE does not make clear how homographs differ from homonyms and homoforms, since it states: “homographs are linguistic elements (words or their forms) written identically in conventional orthography. Lithuanian homographs are usually also homonyms,” e.g.: kasa “braid” – kasa “place for financial transactions” (other lingu-

37

ists consider kasa a true / absolute homonym; this example could be considered a homograph if the present-tense third-person singular forms of the noun kasa and the verb kasti were identical). The singular genitive form varpo of the noun varpas and the present-tense third-person singular form varpo of the verb varpyti are likewise considered homographs (in other linguistic studies, homoforms). LKE also notes that homographs pronounced differently are not uncommon either (e.g.: aušti – aušti, perėjimas – perėjimas, giria – giria) (LKE 1999: 248).

Linguistic studies still lack discussion of cases in which two or more uninflected parts of speech coincide, e.g.: kaip, ir, tik, iki; should these also be classified as homoforms, or should they be considered some other case of ambiguity (the author of this article classifies such cases as homoforms of uninflected parts of speech (see 3. Morphological ambiguity).

3. Morphological polysemy

From the foregoing, it is clear that the terminology presented in linguistic studies varies. To avoid uncertainty about what is actually being analyzed—homoforms, homographs, or partial homonyms—the broader term morphological ambiguity has been chosen. This phenomenon encompasses: 1) inflected and uninflected forms; 2) word forms belonging to different and to the same parts of speech; 3) word forms or words that differ in certain forms or prosodic features, as well as those that are completely identical.

The term ambiguity should be used not only when analyzing homoforms that differ in morphological features, but also in reference to derivational ambiguity (e.g., ištvirkėlis – dobilėlis: the same derivational formant gives rise to different meanings), lexical-semantic ambiguity (e.g., the true homonym žiedas), syntactic-semantic ambiguity (e.g., in the phrase mokytojo pakvietimas, because of the different syntactic relations it is unclear whether the teacher is doing the inviting or is the one being invited), and textual ambiguity (e.g., jis čia išsėdėjo visą vakarą: a broader context is needed to understand the deictic expressions jis and čia).

It must be noted that the analysis of morphological ambiguity is fairly specific: the forms analyzed are written word forms without stress marks

38

forms. The program Lemuoklis, which performs automatic morphological analysis, has no information about the semantics or frequency of use of the words being analyzed; therefore, the words are analyzed without regard to context. As a result, sometimes homoforms are found that are possible only in theory; for example, the form kokio can be morphologically described as the noun kokis, meaning “quality,” and the pronoun koks.

In a morphologically annotated corpus, three types of homoforms are distinguished: 1) forms of inflecting parts of speech, 2) forms of inflecting and uninflecting parts of speech, and 3) forms of uninflecting parts of speech only. The first type of homoforms is illustrated by the following examples: sukalė šuniui b ū d ą – išsiugdė gerą b ū d ą (the form būdą coincides; it may be the singular accusative of the nouns būda or būdas); svarbi p a s l a p t i s – išdavė p ą s l a p t i s (the singular nominative and plural accusative cases of the noun paslaptis coincide). In the second case, for example, the form of a noun, an adverb, and a preposition coincide: atvyko su mano p r i e š a i s – p r i e š a i s matau mišką – stovi p r i e š a i s mane. The particle ir belongs among homoforms of uninflecting parts of speech—it can be a conjunction, particle, or adverb.

As the examples show, words or word forms that are pronounced differently are also considered homoforms. Morphological polysemy, as a phenomenon, encompasses all these types of homoforms. Thus, the automatically and morphologically annotated corpus contains analyses of homoforms, not homographs or homonyms. Since written word forms are analyzed, the focus is primarily on morphological rather than phonetic or semantic differences.

Because of the nature of the analysis, the concept of homoforms has been modified and thus partly encompasses both homographs and partial homonyms. The author of the article believes it is better to choose one clear term rather than several interrelated terms for homonymy. It is hoped that this conception of homoforms will gradually take hold in the analysis of morphologically ambiguous forms.

4. Conclusions

The article examines how linguists use terms related to homonymy and how morphologically ambiguous forms identified through automatic morphological analysis should be named.

39

1. Linguistic works identify several different conceptions of the same term:

1) homoforms and morphological / grammatical homonyms are regarded as either the same thing or as different parts of homonymy;

2) existing definitions of homonymy do not make clear whether homoforms consist of:

a) word forms belonging to the same or different parts of speech,

b) phonetically identical forms of the same word, or forms that may also differ phonetically;

3) some linguists call the same forms homoforms, while others call them partial homonyms;

4) some sources state that homoforms are pronounced identically, while others say they may differ in pronunciation;

5) the criteria for distinguishing homoforms, homonyms, and homographs are defined imprecisely;

6) it is unclear how coincidences of function words are classified.

2. Given the inconsistency of the terminology and the fairly specific nature of the analysis, the term morphological polysemy was chosen, understood as a phenomenon encompassing: 1) inflecting and uninflecting forms; 2) word forms belonging to different and the same parts of speech; 3) word forms or words that differ in certain forms or prosodic elements, as well as those that coincide completely.

3. In the automatically morphologically annotated corpus, more than 47 percent of forms are ambiguous. The term homoforms has been chosen to designate these specific forms. The homoforms of 1) inflecting, 2) inflecting and non-inflecting, and 3) exclusively non-inflecting parts of speech are distinguished; they may differ in prosodic features, i.e., stress placement, pitch accent, or vowel quantity. The term homoforms also encompasses some forms traditionally considered homographs or partial homonyms.

LITERATŪRA

А х м а н о в а О. С. 1986: Словарь омонимов русского языка, Москва.

B a r a u s k a i t ė J. 1979: Leksikologijos pradmenys [Leksinė semantika], Vilnius.

D r o t v i n a s V. 1986: Pagrindinės leksikologijos sąvokos, Vilnius.

J a k a i t i e n ė E. 1980: Lietuvių kalbos leksikologija, Vilnius.

40

Katamba F. 1994: English Words. London and New Your.

KTŽ 1990: K. Gaivenis, S. Keinys. Kalbotyros terminų žodynas, Kaunas.

LK I 1995: Lietuvių kalba, Vilnius.

LKE 1999: Lietuvių kalbos enciklopedija, Vilnius.

Palmer F. R. 1981: Semantics, Cambridge.

Rimkutė E. 2002: Homoformos dabartinės lietuvių kalbos tekstyne. – Lituanistica 2 (50), 86–101.

Соболева П. А. 1980: Словообразовательная полисемия и омонимия, Москва.

Urbutis V. 1956: Lietuvių kalbos leksikos homonimų susidarymo būdai (Filologijos mokslų kandidato disertacijos autoreferatas), Vilnius.

Zinkevičius V. 2000: Lemuoklis – morfologinei analizei. – Darbai ir Dienos 24, 246–273.

HOMONYMS, HOMOFORMS AND HOMOGRAPHS FROM THE VIEWPOINT OF CORPUS LINGUISTICS

Summary

The article deals with a small part of the corpus of the Lithuanian language, which was automatically tagged. The corpus with morphological tags has shown a high level degree of ambiguity of the language: about 47 percent of word forms are ambiguous.

In the Lithuanian linguistics traditionally the greatest attention has been devoted to the lexical homonyms. Other levels of ambiguity have not been analysed. For this reason it was necessary to define the main terms of the morphological ambiguity and to solve the problematic issues of the relevant terminology.

It was difficult to define what are the main differences between homographs, homonyms and homoforms. Traditionally in linguistics it is affirmed that homoforms must coincide phonetically, while homographs are pronounced differently. This article shows that some homoforms can differ by the place of accent, the quality of vowels or intonation. Homoforms were analysed instead of homographs or homonyms, because the main attention was devoted to morphological but not phonetic or semantic differences.

Erika RIMKUTĖ Vytauto Didžiojo universitetas Donelaičio g. 58, Kaunas E. paštas [email protected]

Gauta 2003-03-18

41