Sardinian and English/Structural transfer

From LING073
Revision as of 14:05, 3 May 2023 by Rorlija1 (talk | contribs) (→‎.rtx rules)

Jump to: navigation, search

pre-evaluation

  • WER, PER, and other useful numbers:
 Statistics about input files
 -------------------------------------------------------
 Number of words in reference: 81
 Number of words in test: 80
 Number of unknown words (marked with a star) in test: 14
 Percentage of unknown words: 17.50 %
 Results when removing unknown-word marks (stars)
 -------------------------------------------------------
 Edit distance: 77
 Word error rate (WER): 95.06 %
 Number of position-independent correct words: 12
 Position-independent word error rate (PER): 85.19 %
  • coverage from monolingual transducer on srd.tests:
 Coverage: 97.78%
 Top unknown words in the corpus:
 1	   paìsu
 1	   Lettu
  • coverage from bilingual transducer on srd.tests:
 Number of tokenised words in the corpus: 91
 Coverage: 84.62%
 Top unknown words in the corpus:
 2	   bellu
 1	   paisu
 1	   paìsu
 1	   Lettu
 1	   furatu
 1	   dinari
 1	   punitos
 1	   rica
 1	   datu
 1	   milli
 1	   francos
 1	   betzu
 1	   iscimpru

.rtx rules

  • Item 1: Noun-adjective order in noun phraseology
    Ex:(srd) Sa Frantza est unu paisu bellu. → (eng) France is a beautiful country.
    Outputs for:
    tagger:
    lex
    transfer:
    abc-xyz:

post evaluation

,,