Difference between revisions of "Walloon and English"

From LING073
Jump to: navigation, search
 
(5 intermediate revisions by the same user not shown)
Line 67: Line 67:
 
==Additions==
 
==Additions==
 
* Added 2 structural transfer rules
 
* Added 2 structural transfer rules
−
* ...
+
* Added 3 disambiguation rules
−
* ...
+
** if a possessive clitic is immediately followed by noun or adjective, delete the pronoun reading so only possessive determiner remains
 +
** if that clitic is immediately followed by a finite verb, delete determiner reading so it's kept as a personal pronoun
 +
** when a preposition precedes clitic, again delete the pronoun reading (so it behaves as a determiner inside noun phrase that follows preposition)
 +
* Added 2 lexical selection rules
 +
** for preposition po, when an infinitive verb follows, choose the English translation "to" rather than other senses
 +
** for preposition di, when immediately followed by a noun, select the English translation "of" (overriding "from")
  
 
==Final Evaluation==
 
==Final Evaluation==
 
* Monolingual
 
* Monolingual
−
** Precision:  
+
** Precision: 91.42%
−
** Recall:  
+
** Recall: 30.6% (missing stems, focused predominantly on rules)
−
** Coverage (over wln.corpus.large.txt):
+
** Coverage (over wln.corpus.large.txt): 32.38% (5214 / 16103 tokens)
−
** Words in wln.corpus.large.txt:  
+
** Words in wln.corpus.large.txt: 16,115
−
** Stem Count:  
+
** Stem Count: 131
 
* MT (Walloon -> English)
 
* MT (Walloon -> English)
−
** WER (*.longer.txt):  
+
** WER (*.longer.txt): 45.65%
−
** PER (*.longer.txt):  
+
** PER (*.longer.txt): 36.96%
−
** Correctly Translated Stems (*.longer.txt):
+
** Correctly Translated Stems (*.longer.txt): 46/46
 
** Trimmed Coverage
 
** Trimmed Coverage
−
*** (*.longer.txt):
+
*** (*.longer.txt): 100%
−
*** (*.large.txt):
+
*** (*.large.txt): 25.04%
 
** Number of Tokens
 
** Number of Tokens
−
*** (*.longer.txt):
+
*** (*.longer.txt): 130
−
*** (*.large.txt):
+
*** (*.large.txt): 16,115
  
  

Latest revision as of 22:28, 15 May 2025

Resources developed for Walloon

  • Resourced developed for Walloon

External resources

  • Language Pair Repository: [2]
  • Walloon Transducer: [3]
  • English Transducer: [4]
  • Corpus: [5]

Walloon (wln) → English (eng) Evaluation

Coverage

  • Monolingual Transducer Coverage: 84 / 86 (~0.97674418604651162791)
  • Bilingual Transducer Coverage: 84 / 86 (~0.97674418604651162791)

Sentences

  • (1) wln: "Cwand i l' voeya, Zacariye fourit tot troublé et i hapa sogne."
    • eng: "When Zechariah saw him, he was startled and was gripped with fear."
    • lexical transfer output: 
      ^Cwand<cnjsub>/When<cnjsub>$ ^i<prn><pers><p3><sg>/he<prn><pers><p3><sg>$ ^l'<prn><pers><p3>/him<prn><pers><p3>$ ^voeyér<v><past><p3><sg>/see<v><past><p3><sg>$, ^Zacariye<np><ant><m><sg>/Zechariah<np><ant><sg>$ ^fourî<v><past><p3><sg>/be<v><past><p3><sg>$ ^tot<adv>/all<adv>$ ^troublé<adj><m><sg>/startled<adj><sg>$ ^et<cnjcoo>/and<cnjcoo>$ ^i<prn><pers><p3><sg>/he<prn><pers><p3><sg>$ ^hapar<v><pret><p3><sg>/grip<v><pret><p3><sg>$ ^sogne<n><f><n><m><sg>/fear<n><n><m><sg>$.
    • full translation output: 
      When #he #him #see, Zechariah #be all #startled and #he #grip #fear.
  • (2) wln: "Vosse feme Elizabet vos dinrè on fi, ki vs lomroz Djhan."
    • eng: "Your wife Elizabeth will bear you a son, and you are to call him John."
    • lexical transfer output: 
      ^Vosse<det><poss><p2><pl>/Your<det><pos><p2><pl>$ ^feme<n><f><n><m><sg>/wife<n><n><m><sg>$ ^Elizabet<np><ant><f><sg>/Elizabeth<np><ant><sg>$ ^vos<prn><p2><pl>/you<prn><p2><pl>$ ^dinrer<v><fut><p1><sg>/give<v><fut><p1><sg>$ ^on<det><ind><m><sg>/a<det><ind><sg>$ ^fi<n><m><sg>/son<n><sg>$, ^ki<prn><rel>/whom<prn><rel>$ ^vs<prn><p2><pl>/you<prn><p2><pl>$ ^lomrer<v><fut><p2><pl>/call<v><fut><p2><pl>$ ^Djhan<np><ant><m><sg>/John<np><ant><sg>$.
    • full translation output: 
      #Your #wife Elizabeth #you #give a son, #whom #you #call John.
  • (3) wln: "Djezus houke sès prumîs dissipes."
    • eng: "Jesus calls his first disciples."
    • lexical transfer output: 
      ^Djezus<np><ant><m><sg>/Jesus<np><ant><sg>$ ^houker<v><pres><p3><sg>/call<v><pres><p3><sg>$ ^sès<det><poss><p3><pl>/his<det><pos><p3><pl>$ ^prumîs<adj><m><sg>/first<adj><sg>$ ^dissipe<n><m><pl>/disciple<n><pl>$.
    • full translation output: 
      Jesus #call #his #first disciples.
  • (4) wln: "I ramoennrè bråmint des efants d' Israyel viè l' Diu, leu Signeur."
    • eng: "He will bring back many of the people of Israel to the Lord their God."
    • lexical transfer output: 
      ^I<prn><pers><p3><sg>/He<prn><pers><p3><sg>$ ^ramoenner<v><fut><p1><sg>/bring<v><fut><p1><sg>$ ^bråmint<adv><qnt>/many<adj><qnt>$ ^des<det><ind><pl>/of the<det><ind><pl>$ ^efant<n><m><pl>/people<n><pl>$ ^d'<pr>/of<pr>$ ^Israyel<np><ant><m><sg>/Israel<np><sg>$ ^viè<pr>/to<pr>$ ^l'<prn><pers><p3>/him<prn><pers><p3>$ ^Diu<n><m><sg>/God<np><sg>$, ^leu<det><poss><p3><pl>/their<det><pos><p3><pl>$ ^Signeur<np>/Lord<np>$.
    • full translation output: 
      #He #bring #many #of the peoples of #Israel to #him #God, #their #Lord.
  • (5) wln: "Djezus passéve djondant dèl mér du Galilèye."
    • eng: "Jesus was walking beside the sea of Galilee."
    • lexical transfer output: 
      ^Djezus<np><ant><m><sg>/Jesus<np><ant><sg>$ ^passéver<v><pres><p3><sg>/walk<v><pres><p3><sg>$ ^djondant<pr>/beside<pr>$ ^dèl<pr>/of<pr>$ ^mér<n><m><sg>/sea<n><m><sg>$ ^du<pr>/of<pr>$ ^Galilèye<np>/Galilee<np>$.
    • full translation output: 
      Jesus #walk beside of #sea of #Galilee.
  • (6) wln: "Cwand s' tins d' siervice fourit houte, i ndè rala e s' måjhon."
    • eng: "When his time of service was completed, he returned home."
    • lexical transfer output: 
      ^Cwand<cnjsub>/When<cnjsub>$ ^*s/*s$' ^tin<n><m><pl>/time<n><pl>$ ^d'<pr>/of<pr>$ ^siervice<n><m><sg>/service<n><sg>$ ^fourî<v><past><p3><sg>/be<v><past><p3><sg>$ ^houter<v><pres><p3><sg>/complete<v><pres><p3><sg>$, ^i<prn><pers><p3><sg>/he<prn><pers><p3><sg>$ ^ndè<adv>/thence<adv>$ ^ralar<v><past><p3><sg>/return<v><past><p3><sg>$ ^e<pr>/to<pr>/home<adv>$ ^*s/*s$' ^måjhon<n><f><n><m><sg>/house<n><n><m><sg>$.
    • full translation output: 
      When *s' times of service #be #complete, #he thence #return to *s' #house.
  • (7) wln: "Sol côp, Djezus lès houka."
    • eng: "At once, Jesus called them."
    • lexical transfer output: 
      ^Sol<pr>/On<pr>$ ^côp<n><m><sg>/once<adv><sg>/moment<n><sg>$, ^Djezus<np><ant><m><sg>/Jesus<np><ant><sg>$ ^lès<prn><pers><p3><pl>/them<prn><pers><p3><pl>$ ^houker<v><past><p3><sg>/call<v><past><p3><sg>$.
    • full translation output: 
      On #once, Jesus #them #call.
  • (8) wln: "Po cmincî, l' Univers fourit fait."
    • eng: "In the beginning the Universe was created."
    • lexical transfer output: 
      ^Po<pr>/To<pr>$ ^cmincî<v><inf>/begin<v><inf>$, ^l'<prn><pers><p3>/him<prn><pers><p3>$ ^Univers<np>/Universe<np>$ ^fourî<v><past><p3><sg>/be<v><past><p3><sg>$ ^fait<v><pp>/create<v><pp>$.
    • full translation output: 
      To #begin, #him #Universe #be #create.
  • (9) wln: "Dijhans todi kéke feye."
    • eng: "Or so it would seem."
    • lexical transfer output: 
      ^Dijher<v><pres><p1><pl>/Say<v><pres><p1><pl>$ ^todi<adv>/always<adv>$ ^kéke<det><ind><sg>/some<det><ind><sg>$ ^feye<n><f><n><m><sg>/time<n><n><m><sg>$.
    • full translation output: 
      #Say always #some #time.
  • (10) wln: "Deus des droles di sakîs ås airs di mårticots ont sorviké."
    • eng: "Two of these strange, ape-like creatures survived."
    • lexical transfer output: 
      ^Deus<num><m><pl><num>/Two<num><pl><num>$ ^des<det><ind><pl>/of the<det><ind><pl>$ ^drole<adj><m><pl>/strange<adj><pl>$ ^di<pr>/of<pr>$ ^sakî<n><m><pl>/creature<n><pl>$ ^ås<pr><def>/with<pr>$ ^air<n><m><pl>/air<n><pl>$ ^di<pr>/of<pr>$ ^mårticot<n><m><pl>/ape<n><pl>$ ^ont<v><pres><p3><pl>/have<v><pres><p3><pl>$ ^sorviké<v><pp>/survive<v><pp>$.
    • full translation output: 
      #Two #of the #strange of creatures with airs of apes #have #survive.

Additions

  • Added 2 structural transfer rules
  • Added 3 disambiguation rules
    • if a possessive clitic is immediately followed by noun or adjective, delete the pronoun reading so only possessive determiner remains
    • if that clitic is immediately followed by a finite verb, delete determiner reading so it's kept as a personal pronoun
    • when a preposition precedes clitic, again delete the pronoun reading (so it behaves as a determiner inside noun phrase that follows preposition)
  • Added 2 lexical selection rules
    • for preposition po, when an infinitive verb follows, choose the English translation "to" rather than other senses
    • for preposition di, when immediately followed by a noun, select the English translation "of" (overriding "from")

Final Evaluation

  • Monolingual
    • Precision: 91.42%
    • Recall: 30.6% (missing stems, focused predominantly on rules)
    • Coverage (over wln.corpus.large.txt): 32.38% (5214 / 16103 tokens)
    • Words in wln.corpus.large.txt: 16,115
    • Stem Count: 131
  • MT (Walloon -> English)
    • WER (*.longer.txt): 45.65%
    • PER (*.longer.txt): 36.96%
    • Correctly Translated Stems (*.longer.txt): 46/46
    • Trimmed Coverage
      • (*.longer.txt): 100%
      • (*.large.txt): 25.04%
    • Number of Tokens
      • (*.longer.txt): 130
      • (*.large.txt): 16,115
Retrieved from "http://wikis.swarthmore.edu/ling073/index.php?title=Walloon_and_English&oldid=26296"