Difference between revisions of "Uyghur and English/Ambiguity resolution"

From LING073
Jump to: navigation, search
(→‎Final Ambiguities)
(→‎Final Ambiguities)
Line 37: Line 37:
 
* 3542195 / 3453398 (~1.02571293549)
 
* 3542195 / 3453398 (~1.02571293549)
  
−
Note: I am quite confused at why the change in ambiguity is so small / how it is calculated. Running the tagger on the two example sentences, both in the parallel corpus yields:
+
Note: Running the tagger on the two example sentences, both in the parallel corpus yields:
 
* <code> ^خۇدا<n><nom>$ '''^بۇ<det>$''' ^پاسىل<n><acc>$ ^«<lquot>$^ئاسمان<n><nom>$^»<rquot>$ ^دە<cnjadv>$ ^ئاتى<vblex><pas><p3>$^..<sent>$ </code>
 
* <code> ^خۇدا<n><nom>$ '''^بۇ<det>$''' ^پاسىل<n><acc>$ ^«<lquot>$^ئاسمان<n><nom>$^»<rquot>$ ^دە<cnjadv>$ ^ئاتى<vblex><pas><p3>$^..<sent>$ </code>
 
* <code> ^كېچە<n><nom>$ ^ۋە<cnjcoo>$ ^كۈندۈز<n><nom>$ ^ئۆت<vblex><pas><p3>$ '''^بۇ<prn><dem><nom>$''' ^تۇنجى<adj>$ ^كۈن<n><nom>$ ^بول<vblex><pas><p3>$^..<sent>$ </code>
 
* <code> ^كېچە<n><nom>$ ^ۋە<cnjcoo>$ ^كۈندۈز<n><nom>$ ^ئۆت<vblex><pas><p3>$ '''^بۇ<prn><dem><nom>$''' ^تۇنجى<adj>$ ^كۈن<n><nom>$ ^بول<vblex><pas><p3>$^..<sent>$ </code>
−
This implies that '''at least''' two instances of ambiguity should have been resolved, but seemingly that is not what the test result shows. I find it equally hard to believe that out of the 5 million words in the Wikipedia corpus, 'bu' only appears twice.
 
  
 
==Uyghur and English==
 
==Uyghur and English==

Revision as of 21:36, 8 May 2025

Uyghur

Morphological Ambiguity

These are the ambiguities from coverage-hfst Parallel Corpus

  • coverage: 206 / 215 (~0.95813953488372093023)
  • remaining unknown forms: 9
  • ambiguity: 254 / 215 (~1.18139534883720930233)
  • uig.sentences.txt Wed May 7 06:01:48 PM EDT 2025 206/215 ~0.95813953488372093023 \| 254/215 ~1.18139534883720930233

Large Corpus (wikipedia)

  • coverage: 865241 / 3260897 (~0.26533834095342477852)
  • remaining unknown forms: 2395656
  • ambiguity: 3443129 / 3260897 (~1.05588400982919730369)
  • uig.wikipedia Wed May 7 09:24:25 PM EDT 2025 865241/3260897 ~0.26533834095342477852 \| 3443129/3260897 ~1.05588400982919730369

These are the ambiguities from the shell commands in the assignment:

  • 284 / 270 (~1.05185185185)
  • 3542197 / 3453398 (~1.02571351463)


Identified Ambiguity

  • ^خۇدا/خۇدا<n><nom>$ ^بۇ/بۇ<det>/بۇ<prn><dem><nom>$ ^پاسىلنى/پاسىل<n><acc>$ ^«/«<lquot>$^ئاسمان/ئاسمان<n><nom>$^»/»<rquot>$ ^دەپ/دە<cnjadv>$ ^ئاتىدى/ئاتى<vblex><pas><p3>$^../..<sent>$
  • ^كېچە/كېچە<n><nom>$ ^ۋە/ۋە<cnjcoo>$ ^كۈندۈز/كۈندۈز<n><nom>$ ^ئۆتتى/ئۆت<vblex><pas><p3>$ ^بۇ/بۇ<det>/بۇ<prn><dem><nom>$ ^تۇنجى/تۇنجى<adj>$ ^كۈن/كۈن<n><nom>$ ^بولدى/بول<vblex><pas><p3>$^../..<sent>$

How to resolve ambiguity

The distinction between بۇ ('bu') the pronoun and 'bu' the determiner is defined by whether a noun phrase follows it. We remove the determiner analysis when bu is not followed by a noun phrase. However, since Ugyhur is in SOV order, it is possible that 'bu' as a pronoun subject is followed by a noun phrase. We can select the pronoun 'bu' when it is followed by a noun phrase which is itself followed by a verb phrase. Commonly, this is "'bu' noun verb" or 'bu' adj noun verb.

uig-transducer

Final Ambiguities

These are the ambiguities from the shell commands in the assignment:

Parallel Sentences

  • 283 / 270 (~1.04814814815)

Large Corpus

  • 3542195 / 3453398 (~1.02571293549)

Note: Running the tagger on the two example sentences, both in the parallel corpus yields:

  • ^خۇدا<n><nom>$ ^بۇ<det>$ ^پاسىل<n><acc>$ ^«<lquot>$^ئاسمان<n><nom>$^»<rquot>$ ^دە<cnjadv>$ ^ئاتى<vblex><pas><p3>$^..<sent>$
  • ^كېچە<n><nom>$ ^ۋە<cnjcoo>$ ^كۈندۈز<n><nom>$ ^ئۆت<vblex><pas><p3>$ ^بۇ<prn><dem><nom>$ ^تۇنجى<adj>$ ^كۈن<n><nom>$ ^بول<vblex><pas><p3>$^..<sent>$

Uyghur and English

Retrieved from "http://wikis.swarthmore.edu/ling073/index.php?title=Uyghur_and_English/Ambiguity_resolution&oldid=26082"