Difference between revisions of "Magahi and English"

From LING073
Jump to: navigation, search
(→‎Final Evaluation)
(→‎Final Evaluation)
Line 18: Line 18:
  
 
= Final Evaluation =
 
= Final Evaluation =
−
- Precision and Recall
+
* Precision and Recall
−
:- Precision: 96.58120%
+
** Precision: 96.58120%
−
:- Recall: 83.91089%
+
** Recall: 83.91089%
−
- Coverage: 47.103343307228314662%
+
* Coverage: 47.103343307228314662%
−
- Words in large corpus: 3678603
+
* Words in large corpus: 3678603
−
- Stems in transducer: 359
+
* Stems in transducer: 359
−
- WER over longer: 95.79%
+
* WER over longer: 95.79%
−
- PER over longer: 80.06%
+
* PER over longer: 80.06%
−
- Proportion of stems translated correctly:  
+
* Proportion of stems translated correctly:  
−
- Trimmed coverage
+
* Trimmed coverage
−
:- longer: 80.16%
+
** longer: 80.16%
−
:- large: ??, we don't know have a translation of large
+
** large: ??, we don't know have a translation of large
−
- Number of tokens in
+
* Number of tokens in
−
:- longer: 321
+
** longer: 321
−
:- large: 2,661,822
+
** large: 2,661,822

Revision as of 15:27, 10 May 2021

Resources for machine translation between English and Magahi

External resources

Additions

  • 106 new words
  • 3 new morphological disambiguation rules
  • 3 new transfer rules: objects, adverbs, pronouns

Final Evaluation

  • Precision and Recall
    • Precision: 96.58120%
    • Recall: 83.91089%
  • Coverage: 47.103343307228314662%
  • Words in large corpus: 3678603
  • Stems in transducer: 359
  • WER over longer: 95.79%
  • PER over longer: 80.06%
  • Proportion of stems translated correctly:
  • Trimmed coverage
    • longer: 80.16%
    • large: ??, we don't know have a translation of large
  • Number of tokens in
    • longer: 321
    • large: 2,661,822