Difference between revisions of "Bosnian-Croatian-Montenegrin-Serbian and Macedonian"

From Apertium
Jump to navigation Jump to search
 
(24 intermediate revisions by 2 users not shown)
Line 1: Line 1:
 
{{TOCD}}
 
{{TOCD}}
   
  +
==Source==
=Progress of the work in the bonding period=
 
Insofar, a new dictionary has been started from scratch, some paradigms added from the grammar of croatian, along with some closed word categories. The most extensive work has been done with male noun paradigms, and seems that most work will be done with nouns. Adjectives are inherently more work, but there is less variation.
 
To the dictionary paradigms have been added for verbs, for the present, aorist, imperfect, and futureI tense (the combination of the clitic 'ću' with the infinitive).
 
   
  +
<pre>
==Todo==
 
  +
https://apertium.svn.sourceforge.net/svnroot/apertium/trunk/apertium-sh-mk
  +
</pre>
   
 
==Notes==
;Testing framework
 
   
  +
;Bosnian-Croatian-Montenegrin-Serbian morphological lexicon
* <s>Set up pending/regression tests framework</s>
 
* Set testvoc
 
* Set up corpus/generation-test
 
 
;Serbo-Croatian dictionary
 
 
The reflex of yat:
 
The reflex of yat:
* <s>Adding two additional modes to the monodix (ek/ijek) so that lemmas containing yat can be analysed both as ekavian and ijekavian</s>
+
* Added two additional modes to the monodix (ek/ijek) so that lemmas containing yat can be analysed both as ekavian and ijekavian
* Update the makefile and the xslt machinery so that this works
+
* Updated the makefile and the xslt machinery so that this works
   
 
Verbs(Marked for aspect, transitivity and reflexivity)
 
Verbs(Marked for aspect, transitivity and reflexivity)
* Most of them are from a list extracted from the verbs in the mk monodix (the list is almost completely added, a dosen is missing)
+
* Most of them are from a list extracted from the verbs in the mk monodix
 
* Verb paradigm names are fashioned to contain information about the verb ("aspect_transitivity_paradig/m__vblex"), makes input more convenient
 
* Verb paradigm names are fashioned to contain information about the verb ("aspect_transitivity_paradig/m__vblex"), makes input more convenient
   
Line 25: Line 21:
 
* marked for definiteness
 
* marked for definiteness
 
* entered as quadruples (positive, comparative, superlative, absolute superlative)
 
* entered as quadruples (positive, comparative, superlative, absolute superlative)
 
Nouns:
 
*masculine : a great deal of paradigms covered
 
*feminine : some general cases
 
*neuter : some general cases
 
 
Closed word categories:
 
** <s>prepositions (including the ones of type s/sa and k/ka, which need to be postprocessed in generation)</s>
 
** <s>conjunctions</s>
 
** <s>interjections</s>
 
** particles
 
** pronouns (<s>personal</s>, <s>reflexive</s>, possesive, interrogative, relational, demonstrative (pronoun and adjective), indefinite, negative, ...
 
   
 
Other:
 
Other:
* Add the paradigms from the grammar of Croatian (the one by Barić, Lončarić, Malić, Pavešić, Peti, Zečević, Znika) to the sh monodix [in progress]
+
* Added the paradigms from the grammar of Croatian (the one by Barić, Lončarić, Malić, Pavešić, Peti, Zečević, Znika) to the sh monodix
* Obtain a grammar of Serbian, for reference on differences
 
   
 
;Macedonian dictionary
 
;Macedonian dictionary
   
* Add determiner forms for some pronouns (e.g demonstratives, possessives, etc.) -- things that can modify nouns
+
* Added determiner forms for some pronouns (e.g demonstratives, possessives, etc.) -- things that can modify nouns
  +
* Added some words from closed categories (adverbs, ...)
   
 
;Bilingual dictionary
 
;Bilingual dictionary
* Update the pronoun entries, the symbols in the monodix have been adjusted to correspond more closely to the analysis in the macedonian monodix
+
* Updated the pronoun entries, the symbols in the monodix have been adjusted to correspond more closely to the analysis in the macedonian monodix
   
 
;Transfer rules
 
;Transfer rules
  +
  +
==Transfer rules==
  +
Three stage transfer is used.
  +
  +
Some problems:
  +
  +
Genitive constructions, problem in distinguishing partitive from possesive:
  +
* Čaša vode = Чаша вода [partitive]
  +
* Pilotkinje Vojske Srbije == Пилотките на Српската војска [possesive]
  +
* TODO: After the lexicon in the analyser becomes sufficiently large grep out all "(Noun) (Noun + Genitive)" occurences in i.e. the wikipedia corpus, and find pairs which:
  +
** '''More frequently appear as partitive'''
  +
** '''More frequently appear as possesive'''
  +
** Can appear either way
  +
and afterwards ignore the latter case, and make two categories of noun pairs.
  +
  +
Tricky sequences (instrumental adjectives):
  +
* "...upravljanje '''velikom''', '''jakom''' ''''pticom'''' " == "...управување '''со''' '''голема''', '''силна''' ''''птица'''' ".
  +
Only one preposition needed for the whole chain, which can be of any length.
  +
*:What we do here is just make long chunks in the t1x, e.g. in this case: ADJ ADJ CM ADJ NOM
  +
*:These sequences will probably be fairly infrequent though... - [[User:Francis Tyers|Francis Tyers]] 09:05, 20 July 2011 (UTC)
  +
  +
Change of gender in translation:
  +
...vožnja'''<f>''' zrakoplovom bila je odlučujuća'''<f>'''... == ...возење'''<nt>''' со авион беше решавачко'''<f>'''.
  +
Potentialy too far to be matched.
  +
  +
Obligatory clitic with definite object
  +
* I saw the man == Го видов човекот
  +
The clitics must preceed the finite verb, in this order:
  +
* subjunctive-negative-mood-aux-ethical dative-dative object-accusative object-verb
  +
* ...да не ќе сум си му го дал... == that I won't have given it to him...
  +
Possesive genitive to на construcions:
  +
* Anina čaša == чашата на Ана (Ana's glass)
  +
* Anina ruka == раката на Ана (Ana's hand)
  +
* (possible TODO in the mk analyser: add an analysis for possesive adjectives, currently doesn't exist)
  +
Preposition 's' when standing with genitive does not translate as 'со':
  +
* ...sam poletjela s avionske piste...==...полетав '''од''' авионската писта... and not ...'''со''' авионската писта..., which would have an instrumental meaning
   
 
==Disambiguation rules==
 
==Disambiguation rules==
For all cases modifiers (adjectives, numbers, pronouns transfer number, case and gender)
+
For all cases modifiers (adjectives, numbers, pronouns) transfer number, case and gender, so instances like these were used to disambiguate
  +
* ...sa njenim toplim nježnim rukama - A sequence in instrumental
  +
* ...sa svojih toplih nježnih ruku - A sequence in genitive
   
 
Nominative or accusative:
 
Nominative or accusative:
Line 76: Line 98:
   
 
Numbers:
 
Numbers:
* Numbers 2-4 govern noun phrases differently (remnants of dual)
+
* Numbers 2-4 govern noun phrases differently (remnants of dual), two variants:
  +
** "s trima lijepim ženama" (with three women). The number and the rest of the phrase take the dual forms (Nom,Acc,Voc==Gen.Sg, Gen=>Gen.Pl (triju lijepih žena), Dat,Loc,Ins=>Ins.Pl (trima lijepim ženama) ). This variant is more literary.
** Number takes the head case of the phrase, the modified phrase takes the dual cases
 
  +
** "s tri lijepe žene" (the number is in a frozen form, and the rest of the phrase gets genitive, the actual meaning is determined from context or prepositions). This variant is closer to actual speech.
 
* Remove 'Num' reading of 'sto' (N or Num) if preceeded by a determiner.
 
* Remove 'A' / 'Det' following noun.
 
* Remove 'V' after possessive pronoun.
 
* Remove 'Pcle' for 'ovaj' if not followed by a comma.
 
* Remove 'Adj' if 'Adj' OR 'Prop' if previous word is 'U' and second word not part of an NP (e.g. pronoun, finite verb, ...)
 
** ''U Bugarskoj je do sada dijagnosticirano samo 366 slučajeva, a u Hrvatskoj 341.''
 
   
 
==See also==
 
==See also==
Line 90: Line 106:
 
* [[/Pending tests|Pending tests]]
 
* [[/Pending tests|Pending tests]]
 
* [[/Regression tests|Regression tests]]
 
* [[/Regression tests|Regression tests]]
  +
* [[/Final_report|Final report]]
  +
  +
;TODOs
 
* Set up corpus/generation-test
   
 
==External links==
 
==External links==
Line 97: Line 117:
 
* [http://rechnik.on.net.mk/ Macedonian<->Serbian online dictionary]
 
* [http://rechnik.on.net.mk/ Macedonian<->Serbian online dictionary]
 
* [http://www.makedonski.info/ Word definitions for Macedonian]
 
* [http://www.makedonski.info/ Word definitions for Macedonian]
  +
* [http://www.mling.ru/iazik/makedonski/history_makedonski.pdf Блаже Конески - Историја на македонскиот јазик (Blaže Koneski - History of the Macedonian Language; in Macedonian, cyrilic)]
   
 
==Further reading==
 
==Further reading==
Line 106: Line 127:
 
* SEELRC [http://www.seelrc.org:8080/grammar/pdf/stand_alone_bcs.pdf Bosnian / Croatian / Serbian Reference Grammar]
 
* SEELRC [http://www.seelrc.org:8080/grammar/pdf/stand_alone_bcs.pdf Bosnian / Croatian / Serbian Reference Grammar]
 
* SEELRC [http://www.seelrc.org:8080/grammar/pdf/stand_alone_macedonian.pdf Macedonian Grammar]
 
* SEELRC [http://www.seelrc.org:8080/grammar/pdf/stand_alone_macedonian.pdf Macedonian Grammar]
[[Category:Serbo-Croatian and Macedonian|*]]
+
[[Category:Bosnian-Croatian-Montenegrin-Serbian and Macedonian|*]]
  +
[[Category:Bosnian-Croatian-Montenegrin-Serbian]]
  +
[[Category:Macedonian]]

Latest revision as of 05:34, 22 August 2017

Source[edit]

https://apertium.svn.sourceforge.net/svnroot/apertium/trunk/apertium-sh-mk

Notes[edit]

Bosnian-Croatian-Montenegrin-Serbian morphological lexicon

The reflex of yat:

  • Added two additional modes to the monodix (ek/ijek) so that lemmas containing yat can be analysed both as ekavian and ijekavian
  • Updated the makefile and the xslt machinery so that this works

Verbs(Marked for aspect, transitivity and reflexivity)

  • Most of them are from a list extracted from the verbs in the mk monodix
  • Verb paradigm names are fashioned to contain information about the verb ("aspect_transitivity_paradig/m__vblex"), makes input more convenient

Adjectives:

  • marked for definiteness
  • entered as quadruples (positive, comparative, superlative, absolute superlative)

Other:

  • Added the paradigms from the grammar of Croatian (the one by Barić, Lončarić, Malić, Pavešić, Peti, Zečević, Znika) to the sh monodix
Macedonian dictionary
  • Added determiner forms for some pronouns (e.g demonstratives, possessives, etc.) -- things that can modify nouns
  • Added some words from closed categories (adverbs, ...)
Bilingual dictionary
  • Updated the pronoun entries, the symbols in the monodix have been adjusted to correspond more closely to the analysis in the macedonian monodix
Transfer rules

Transfer rules[edit]

Three stage transfer is used.

Some problems:

Genitive constructions, problem in distinguishing partitive from possesive:

  • Čaša vode = Чаша вода [partitive]
  • Pilotkinje Vojske Srbije == Пилотките на Српската војска [possesive]
  • TODO: After the lexicon in the analyser becomes sufficiently large grep out all "(Noun) (Noun + Genitive)" occurences in i.e. the wikipedia corpus, and find pairs which:
    • More frequently appear as partitive
    • More frequently appear as possesive
    • Can appear either way

and afterwards ignore the latter case, and make two categories of noun pairs.

Tricky sequences (instrumental adjectives):

  • "...upravljanje velikom, jakom 'pticom' " == "...управување со голема, силна 'птица' ".

Only one preposition needed for the whole chain, which can be of any length.

  • What we do here is just make long chunks in the t1x, e.g. in this case: ADJ ADJ CM ADJ NOM
    These sequences will probably be fairly infrequent though... - Francis Tyers 09:05, 20 July 2011 (UTC)

Change of gender in translation: ...vožnja<f> zrakoplovom bila je odlučujuća<f>... == ...возење<nt> со авион беше решавачко<f>. Potentialy too far to be matched.

Obligatory clitic with definite object

  • I saw the man == Го видов човекот

The clitics must preceed the finite verb, in this order:

  • subjunctive-negative-mood-aux-ethical dative-dative object-accusative object-verb
  • ...да не ќе сум си му го дал... == that I won't have given it to him...

Possesive genitive to на construcions:

  • Anina čaša == чашата на Ана (Ana's glass)
  • Anina ruka == раката на Ана (Ana's hand)
  • (possible TODO in the mk analyser: add an analysis for possesive adjectives, currently doesn't exist)

Preposition 's' when standing with genitive does not translate as 'со':

  • ...sam poletjela s avionske piste...==...полетав од авионската писта... and not ...со авионската писта..., which would have an instrumental meaning

Disambiguation rules[edit]

For all cases modifiers (adjectives, numbers, pronouns) transfer number, case and gender, so instances like these were used to disambiguate

  • ...sa njenim toplim nježnim rukama - A sequence in instrumental
  • ...sa svojih toplih nježnih ruku - A sequence in genitive

Nominative or accusative:

  • Select Accusative if there's a preceeding Accusative preposition
  • Select Accusative if there's a preceeding transitive verb (direct object rule)

Genitive or accusative:

  • Disambiguate based on prepositions (the intersection is only "u")
  • Select accusative if there's a preceeding transitive verb (direct object rule)

Dative or locative:

  • Select dative if there's no preceeding preposition or modifier in dative
  • Select dative if there's a preceeding dative preposition
  • Select locative if there's a preceeding locative preposition

Instrumental (unambiguous in singular):

  • Select instrumental if there's a preceeding instrumental preposition
  • In plural identical to Dative/Locative
    • (easily disambiguated from Locative, since the latter is entirely prepositional)
    • Possible problems in the plural with instrumental/non-instrumental, though not yet encountered
      • "Ljudima sam pomeo pod" - 'To the people' or 'using the people'
      • "Ploviti morima" - To sail 'to the seas' or 'across seas' (or 'using the seas')

Numbers:

  • Numbers 2-4 govern noun phrases differently (remnants of dual), two variants:
    • "s trima lijepim ženama" (with three women). The number and the rest of the phrase take the dual forms (Nom,Acc,Voc==Gen.Sg, Gen=>Gen.Pl (triju lijepih žena), Dat,Loc,Ins=>Ins.Pl (trima lijepim ženama) ). This variant is more literary.
    • "s tri lijepe žene" (the number is in a frozen form, and the rest of the phrase gets genitive, the actual meaning is determined from context or prepositions). This variant is closer to actual speech.

See also[edit]

TODOs
  • Set up corpus/generation-test

External links[edit]

Further reading[edit]

  • Ivan Todorović "Disambiguation of Serbian sentences with Unitex".

References[edit]