Забыли пароль?
Ещё не зарегистрированы? Регистрация
RSS-ленты новостей
rss20.gif

Портал был создан при финансовой поддержке Российского гуманитарного научного фонда (РГНФ), проект № 07-04-12140в.

Портал зарегистрирован 05 августа 2010 г. в Федеральной службе по надзору в сфере связи, информационных технологий и массовых коммуникаций (Роскомнадзор) в качестве средства массовой информации, номер свидетельства ЭЛ № ФС 77 - 41581. Учредитель В. А. Баранов.

(c) "Информационные технологии и письменное наследие", 2008-2017

PRELIMINARY CONSIDERATIONS CONCERNING THE AUTOMATED LEMMATISATION OF MIDDLE BULGARIAN TEXTS PDF Печать E-mail
Автор(ы): Juergen Fuchsbauer   
27.08.2012 г.
Summary. The present paper attempts to define the philological preconditions for the digital processing of texts written in Middle Bulgarian with the help of software applicable for other recensions of Church Slavonic. With the Slavonic Dioptra as an example, required adaptations on the levels of graphetics, graphematics, and morphology are discussed.

The Dioptra is a voluminous Greek didactic poem composed as a dialogue of body and soul, which was translated into Middle Bulgarian Church Slavonic around the middle of the fourteenth century. As was first noted by Franz von Miklosich, it contains an abundance of remarkable lexical material, which until now has not been analysed conclusively. Therefore, the bilingual critical edition being currently prepared at Vienna University shall be completed by a dictionary eventually disclosing the lexicon of the poem. In view of its considerable length—the Dioptra consists of approx. 62.000 words—a largely automated lemmatisation appears highly desirable. This requires a device for approximate string matching directly applicable to Middle Bulgarian texts, which, as to my knowledge, for now does not exist. The present paper lists the deviations of the Dioptra from Old Church Slavonic relevant to the automated processing of the text. Its goal is to outline from a philological point of view the prerequisites for an adaptation of approximate string matching techniques developed for other variants of Slavonic[1] to the Dioptra. At that, OCS is unquestionably a more natural point of reference than Old Russian. The results can be expected to be applicable for other Middle Bulgarian texts as well.

Our edition relies on the L’viv manuscript of the Dioptra (LNB NAN imeni Stefanyka MV-418), as this is the only completely preserved Middle Bulgarian testimony of the poem. First of all, in order to allow fuzzy string matching, the software processing the text should be capable of abstracting from certain graphic peculiarities of the ms represented in the print version. Thus, the 12 letters (out of a total of 51 used in our edition) representing positional or arbitrary allographs should be assigned to the superordinate graphemes (2 and ¬ to е;[2] s to ł; andto и;[3] w, 3, and 5 to о;[4] У and ? to ;[5] û to ¥; and v to y.[6]). Additionally, the lemmata in the dictionary should appear in a corresponding “abstract” form, relieving the reader of some time-consuming guesswork. Of course, the actual spelling is to be preserved in the single entries listed under the respective headwords.

I do not expect the operations necessary for a simplification of that kind to cause much trouble. By contrast, the frequent alternations of graphemes resulting from phonetic shift can be assumed to pose a much bigger challenge both to computational scientists entrusted with the task of adapting existing software to the requirements of Middle Bulgarian, and to philologists processing the data thus gained. I examined the spelling principles of the Middle Bulgarian Dioptra mss in a recent paper in detail;[7] therefore, I shall only give a brief overview here.

Following graphematic alternations appear regularly in the L’viv ms of the Dioptra (and, of course, in many other Middle Bulgarian mss):

ł ~ з / з ~ ł: only a few cases contradict the etymological spelling; most of these deviations seem to be lexicalised (e.g., the adjective полезн¥и, is always spelt with з, the noun полłа, by contrast, unexceptionally with ł).

л ~ ø      epenthetic l is comparatively frequently omitted.

ъ ~ ø (/ ü / о): weak ъ may be skipped, but is usually preserved in spelling; it is hardly ever replaced by ü; о-vocalism occurs only in a few words (любовü, нач­токъ) and seems to be lexicalised.

¥ ~ и:     both are only exceptionally mistaken for one another; a few cases of regular, lexicalised commutation occur (нинэ, посилати).

ь ~ ø / е / ъ: weak ü may be skipped or replaced by ъ, but is usually preserved; strong historic ü appears as е, weak ü vocalised in order to split consonant clusters either as ü or ъ.

э ~ я:    a complementary distribution prevails; э is used after soft consonants, я at the word onset and at morpheme boundaries; after vowels only а appears.

­ ~ ©:     as a rule, the choice of one of the nasal graphemes is influenced, but not strictly determined, by the quality of the preceding sound; ­ is preferred at the word onset, after soft consonants and forward vowels, © after hard consonants with a more ambiguous distribution after sibilants and non-forward vowels.

In general, the spelling of the L’viv ms of the Dioptra seems to be fairly consistent and highly lexicalised. Words deviating from a presupposed OCS standard are likely to be spelt in the same way in other occurrences as wellthough the total number of possible variations is rather high, only a limited set is realised. This can, once appropriate parameters were defined, be expected to facilitate approximate string matching significantly.

A pivotal point in the automated processing of a text is evidently the correct assignment of inflexion forms. In the following, I give an overview of the desinences present in the Dioptra which do not or not regularly occur in OCS (merely graphematic phenomena covered above are not quoted expressly; e.g. землэ = nom. sg. fem. ja-stem). For comparison I used [Diels, 1963]. Most of these endings are all but uncommon in Middle Bulgarian; not a few occur even sporadically in OCS (those mentioned by Diels are given in italics).

-а  nom. sg. fem. and masc. former ī-stems, which were adopted to the ja-stem-paradigm (млъниа, с©диа)

-е   nom. sg. masc. jo-stems: proper names ending in -ιος in Greek (e.g. григорие)

      nom. sg. neutr. of the short form of the part. praet. act. (и дрэво е­ ветхо же и изгнивъше; according to [Diels, 1963: 242], also attested in Supr.)

     acc. sg. of r-st. (матере, дъωере; according to [Diels, 1963: 178], also in Sav. and Supr.)

      nom. pl. of some masc. jo-stems (коне, коваче, прэлþбодэе)

-еве nom. pl. of monosyllabic masc. jo-stems (rare! e.g. врачеве, плачеве, краеве; cf. [Diels, 1963: 159])

-еи loc. fem. long form of soft adjectives (rare! въ послэднеи старости; въ прочеи твари)

      gen. pl. of masc. jo-stems (e.g. м©жеи; cf. [Diels, 1963: 159])

-емъ    loc. sg. masc./neutr. of the long form of soft adjectives, comparatives, and part. praes./praet. act. (въ насто©ωемъ житии)

-ехъ    loc. pl. of masc./neutr. jo-stems (въ агньцехъ)

-ие  nom. pl. masc. of jo-stems, especially of those ending in -tel’, -ar’, and soft monosyllabic roots (e. g. родителие, р¥барие, царие, м©жие)

-ии gen. pl. of masc. jo-stems (м©жии)

-м¥    1. pers. pl. of the athemat. verbs (есм¥, вэм¥, имам¥, дам¥; according to [Ivanova-Mirčeva and Charalampiev, 1999: 134], this ending is already attested in OCS documents)

-ове nom. pl. of monosyllabic masc. o-stems (e.g. родове; cf. [Diels, 1963: 156])

-омоy   dat. sg. masc./neutr. of the long form of hard adjectives, comparatives, part. praes. act., praes. pass., praet. act., praet. pass. (e. g. богатэ©ωомоy)

-омъ    instr. sg. and dat. pl. of neutr. jo-stems ending in -ie in nom. sg. (искоyшениомъ зъмииноN и зависти­ диаволе­); rarely also of masc. with a stem ending in a vowel (after the loss of intervocalic j; e.g. къ садоyкеомъ, къ иоyдеомъ)

     loc. sg. masc./neutr. of the long form of hard adjectives (въ четврътомъ словэ) and the part. praes./praet. pass. (въ ... насажденомъ раи)

-охъ  loc. pl. of masc./neutr. o-stems (въ нэдрохъ; masc. already in OCS, cf. [Diels, 1963: 157])

-(ü)ми instr. pl. of masc. jo-stems (оyч·телми; according to [Diels, 1963: 157], -ъми is attested with OCS o-stems)

         instr. pl. of the neutr. jo-stems ending in -ie in nom. sg. (wUвэωанми)

-эмъ instr. (!) sg. masc./neutr. of hard adjectives (съ шоyмомъ велицэмъ; otherwise also as regular loc. form)

-­     nom./acc. pl. of r-stems (дъωер­)

       acc. pl. of masc. n-stems (степен­)

-©(и) nom. sg. masc. short (long) form of the part. praes. act. replacing -¥(и)

Many of these morphological innovations, which affected almost exclusively the nominal and adjectival inflexion, were caused by inter-paradigmatic equalisation.[8] Therefore, most of the respective desinences should be readily identifiable for software applicable to OCS as they appear in an either identical or similar form in at least one other paradigm (e.g. -üми in the ĭ-stems, -ове in the former ŭ-stems). On the other hand, intra-paradigmatic neutralisation (as in -эмъ for the instr. sg. of masculine and neuter adjectives) is not common enough to seriously aggravate the problem of homonymy, which can be expected to leave the editor with a lot of manual work anyway.

All in all, despite the loss of the casus in the contemporary vernacular, in respect to morphology the Dioptra preserved an artificial standard close to OCS. Therefore a digital processing of the poem does not seem less promising than the processing of OCS or Old Russian texts.


[1] I have in mind the OldEd developed at Izhevsk State Technical University.

[2] The letter 2 is preferred after vowels, at the word onset, and at the end of lines, but may occur in any position; ¬ appears only in ¬T΅ (= ¬стъ) and, occasionally, in ¬„ωе.

[3] The letter ∙ is frequently, yet not obligatorily, used in front of vowels, but may appear in any position; ¶ is restricted to Greek loanwords (¶„ндктиwн, ¶„2реи) and names of Greek or Hebrew origin (¶„ппократъ, ¶„2„зекилъ).

[4] Both w and 3 may appear in any position; w is clearly preferred at the word onset; 5 is notoriously restricted to the word oko.

[5] Digraphic № is by far most common, but may be replaced by У in any position; ? (an v set above an о) occurs only exceptionally.

[6] The other letters, s, û, and v, exceptionally replace their more frequent counterparts.

[7] Remarks on the Grammar of the Slavonic Dioptra. Part I: Orthography and Phonetics” (submitted for the 2012 issue of Scripta & e-Scripta).

[8] We detect a few more isolated instances of unproductive stems adopting desinences of their productive counterparts, that were not incorporated in the list above: n-stem дüнü at least twice took over jo-stem endings (gen. sg. и днэ не вэси, otherwise: дне оного; also: dat. sg. днþ), ū-stem црüк¥ the a-stem acc. pl. -¥ (цръкв¥ łиздати), and мати the dat. pl. ja-stem -эмъ (матерэмъ).

 
« Пред.   След. »