Difference between revisions of "Munsee/Transducer"

From LING073
Jump to: navigation, search
(→‎Generator Evaluation)
Line 37: Line 37:
 
* Current coverage over corpus: 40/1034 (~3.9%)
 
* Current coverage over corpus: 40/1034 (~3.9%)
 
* Number of gold matches for the generation test: 0 :(
 
* Number of gold matches for the generation test: 0 :(
 +
 +
=== Final Evaluation of Morphological Generation ===
 +
* Number of passing tests:
 +
* Number of failing tests:
 +
* Number of twol rules added:
 +
* Current coverage over corpus:
  
 
[[Category:Sp25_Transducers]] [[Category: Munsee]]
 
[[Category:Sp25_Transducers]] [[Category: Munsee]]

Revision as of 20:29, 16 March 2025

Link to Transducer Repository

Analyser Evaluation

  • Total number of stems in transducer: 20
  • Current coverage over combined corpus: 33/1024
  • Current list of top unknown words:
    • uch
    • waak
    • Nal
    • weemi
    • ha
    • mah
    • ang
    • ngumee
    • uw
    • nal
    • taa
    • nzhiis
    • nguk
    • kway
    • wtulaawal
    • eet
    • wa
    • njihnal
  • Number of tests that pass in each test: 61/73 pass (0% match gold, likely due to transducer outputting multiple possible forms, some of which are not the "gold" one)

Notes

  • Coverage increased from 11/1024 to 33/1024 with the addition of 'Noah' and 'Pahtamawaas' (God) as animate noun stems.

Generator Evaluation

Initial Evaluation of Morphological Generation

  • Number of passing morphological analysis tests: 34/82 (~41.5%)
  • Number of failing morphological analysis tests: 48/82
  • Current coverage over corpus: 40/1034 (~3.9%)
  • Number of gold matches for the generation test: 0 :(

Final Evaluation of Morphological Generation

  • Number of passing tests:
  • Number of failing tests:
  • Number of twol rules added:
  • Current coverage over corpus: