Urum and English/Structural Transfer

From LING073
Revision as of 17:27, 29 April 2023 by Scasada1 (talk | contribs) (→‎(UUM ‣ ENG) Transfer Evaluation)

Jump to: navigation, search

External resources

Transducer Links

Language Pair Repository

Structural Transfer Pre-Evaluation

Statistics About Input Files

(ling073-uum-eng-corpus/eng.sentences.txt & ling073-uum-eng/pre-eval_uum-eng.tests.txt)
Number of words in reference: 231
Number of words in test: 182
Number of unknown words (marked with a star) in test: 76
Percentage of unknown words: 41.76 %


Results when removing unknown-word marks (stars)

Edit distance: 231
Word error rate (WER): 100.00 %
Number of position-independent correct words: 7
Position-independent word error rate (PER): 96.97 %

Coverage Tests

TOP UNKNOWN WORDS:
     3 ^да/*да$
     2 ^тʼетти/*тʼетти$
     2 ^тʼетирдилер/*тʼетирдилер$
     2 ^доғуштан/*доғуштан$
     1 ^ӧстӱ/*ӧстӱ$
     1 ^җеннет/*җеннет$
     1 ^эң/*эң$
     1 ^эништилер/*эништилер$
     1 ^энди/*энди$
     1 ^ышхын/*ышхын$
     1 ^ышихындан/*ышихындан$
     1 ^чобанлара/*чобанлара$
     1 ^чобанлар/*чобанлар$
     1 ^чобаллара/*чобаллара$
     1 ^чине/*чине$
     1 ^хырды/*хырды$
     1 ^хутхарды/*хутхарды$
     1 ^хорхундулар/*хорхундулар$
     1 ^хонуштулар/*хонуштулар$
     1 ^хонду/*хонду$
coverage: 130 / 206 (~0.63106796116504854369)
remaining unknown forms: 76

(UUM ‣ ENG) Transfer Evaluation

Example 1 (Verb Pronoun Marking)

  • (uum) алйым → (eng) I find
    (uum) ал<v><tv><pres><p1><sg> → (eng) I<prn><subj><p1><m><sg>find<vblex><pres>
Tagger Output:
^ал<v><tv><pres><p1><sg>$^.<sent>$

Example 2 (Verb Past-Tense)

  • (uum) булду → (eng) he found
    (uum) бул<v><tv><past><p3><sg> → (eng) he<prn><p3><m><sg>find<vblex><past>

Example 3 (Verb Negativity)

  • (uum) булмады → (eng) he did not find
    (uum) бул<v><tv><neg><past><p3><sg> → (eng) he<prn><p3><m><sg>did<vblex><past>not<adv> find<vblex><pres>

Example 4 (Ablative Case)

  • (uum) кӧктэн → (eng) from sky
    (uum) кӧк<n><abl> → (eng) from<pr>sky<n><sg>

Example 5 (Locative Case)

  • (uum) бахчыда → (eng) in sky
    (uum) бахчы<n><loc> → (eng) in<pr>sky<n><sg>

Structural Transfer Post-Evaluation

Statistics About Input Files

Number of words in reference: 231
Number of words in test: 236
Number of unknown words (marked with a star) in test: 76
Percentage of unknown words: 32.20 %

Results when removing unknown-word marks (stars)

Edit distance: 238
Word error rate (WER): 103.03 %
Number of position-independent correct words: 35
Position-independent word error rate (PER): 87.01 %

Coverage Tests

TOP UNKNOWN WORDS:
     3 ^да/*да$
     2 ^тʼетти/*тʼетти$
     2 ^тʼетирдилер/*тʼетирдилер$
     2 ^доғуштан/*доғуштан$
     1 ^ӧстӱ/*ӧстӱ$
     1 ^җеннет/*җеннет$
     1 ^эң/*эң$
     1 ^эништилер/*эништилер$
     1 ^энди/*энди$
     1 ^ышхын/*ышхын$
     1 ^ышихындан/*ышихындан$
     1 ^чобанлара/*чобанлара$
     1 ^чобанлар/*чобанлар$
     1 ^чобаллара/*чобаллара$
     1 ^чине/*чине$
     1 ^хырды/*хырды$
     1 ^хутхарды/*хутхарды$
     1 ^хорхундулар/*хорхундулар$
     1 ^хонуштулар/*хонуштулар$
     1 ^хонду/*хонду$
coverage: 130 / 206 (~0.63106796116504854369)
remaining unknown forms: 76