Difference between revisions of "Rungus/Transducer"

From LING073
Jump to: navigation, search
(→‎Final Evaluation of Morphological Generation)
(→‎Final Evaluation of Morphological Generation)
Line 54: Line 54:
  
 
Our new coverage is: 439 / 1226 (~35.8075%)
 
Our new coverage is: 439 / 1226 (~35.8075%)
 +
*Command: <code>coverage-hfst drg.corpus.basic.txt drg.automorf.hfst</code>
  
 
[[Category:Rungus]]
 
[[Category:Rungus]]
 
[[Category:Sp25 Transducers]]
 
[[Category:Sp25 Transducers]]

Revision as of 20:33, 7 March 2025

GitHub [1]

Evaluation

Number of Stems: 100

Coverage: 416/1226 (~33.9314%)

Current List of Unknown Words:

  • ong
  • no
  • di
  • tu
  • om
  • nu
  • ogi
  • ku
  • oku
  • an
  • amu
  • dau
  • savo
  • po
  • ko
  • agai
  • dikou
  • Amu
  • vokon

All tests in each of the corpus pass (I think?), but not many match gold. In fact, only 29 of them match gold.

Generator Evaluation

Initial Evaluation of Morphological Generation

I believe all tests passed initially.

Current Coverage: 416 / 1226 (~33.9148%)

  • Command: coverage-hfst drg.corpus.basic.txt drg.automorf.hfst

Current Gold Matches for Generation Tests: apertium-regtest test -c .-gen

  • Intransitive Prefixes: 3/6 (50.0%) match gold
  • Pluralizing Nouns: 0/6 (0.0%) match gold
  • Collectivizing Nouns: 0/3 (0.0%) match gold
  • Personal Pronouns: 8/16 (50.0%) match gold
  • Demonstrative Determiners: 6/6 (100.0%) match gold
  • Demonstrative Pronouns: 12/13 (92.31%) match gold
  • Contractions: 0/6 (0.0%) match gold
  • Reduplication: 0/6 (0.0%) match gold
  • Enclitics: 0/9 (0.0%) match gold
  • Transitive Prefixes: 0/23 (0.0%) match gold

Final Evaluation of Morphological Generation

note down the number of passing and failing tests after adding the morphological generator

We previously did not have any twol rules. After adding the morphological generator, we have 14 twol rules.

Our new coverage is: 439 / 1226 (~35.8075%)

  • Command: coverage-hfst drg.corpus.basic.txt drg.automorf.hfst