Difference between revisions of "Amis/Transducer"
| Line 103: | Line 103: | ||
We added 5 twol rules. | We added 5 twol rules. | ||
| − | + | Number of tokenised words in the corpus: 225 | |
| − | + | Coverage: 30.67% | |
| − | + | Top unknown words in the corpus: | |
| − | + | 13 ko | |
| − | + | 11 no | |
| − | + | 10 tamdaw | |
| − | + | 9 to | |
| − | + | 7 i | |
| − | + | 7 o | |
| − | + | 4 caay | |
| − | + | 4 tatiihay | |
| − | + | 3 rakat | |
| − | + | 3 sa | |
| − | + | 2 O | |
| − | + | 2 ciraraway | |
| − | + | 2 Kawas | |
| − | + | 2 Rikec | |
| − | + | 2 ka | |
| − | + | 2 mo | |
| − | + | 2 ^celay | |
| − | + | 2 Nawiro | |
| − | + | 2 kasakitakit | |
| − | + | 1 so | |
| − | + | Translation time: 0.05767536163330078 seconds | |
| − | |||
[[Category:Amis]] [[Category:Sp23_Transducers]] | [[Category:Amis]] [[Category:Sp23_Transducers]] | ||
Revision as of 14:47, 14 March 2023
Contents
Evaluation
Code: https://github.swarthmore.edu/Ling073-sp23/ling073-ami
There are 75 stems in the transducer.
TOP UNKNOWN WORDS:
115 ^ko/*ko$
95 ^i/*i$
58 ^to/*to$
55 ^no/*no$
43 ^o/*o$
34 ^tamdaw/*tamdaw$
24 ^Tapang/*Tapang$
23 ^takowanan/*takowanan$
17 ^ato/*ato$
17 ^Tapangaw/*Tapangaw$
16 ^tatiihay/*tatiihay$
14 ^Tisowanan/*Tisowanan$
13 ^O/*O$
12 ^ada/*ada$
12 ^Kawas/*Kawas$
11 ^ma/*ma$
10 ^mo/*mo$
10 ^ka/*ka$
10 ^faloco/*faloco$
10 ^dmak/*dmak$
coverage: 759 / 2107 (~0.36022781205505457997) remaining unknown forms: 1348 ambiguity: 2366 / 2107 (~1.12292358803986710963) ami.corpus Thu Mar 2 09:39:48 PM EST 2023 759/2107 ~0.36022781205505457997 \| 2366/2107 ~1.12292358803986710963
Corpus 1 of 5: Instrumental Applicative-morph
6/6 (100.0%) tests pass (6/6 (100.0%) match gold)
Corpus 2 of 5: Locative Applicative-morph
6/6 (100.0%) tests pass (6/6 (100.0%) match gold)
Corpus 3 of 5: Factual Mood-morph
5/5 (100.0%) tests pass (5/5 (100.0%) match gold)
Corpus 4 of 5: Pronouns-morph
24/24 (100.0%) tests pass (11/24 (45.83%) match gold)
Corpus 5 of 5: Number Plural Marker-morph
3/3 (100.0%) tests pass (2/3 (66.67%) match gold)
All tests pass.
Notes
After adding the most uncommon word a<conj> to the transducer, coverage went from 19.93% to 27.25%. After accomodating for o/u alternation, coverage went from 27.25% to 36.02%.
Generator Evaluation
Initial evaluation of morphological generation
number of passing and failing morphological analysis tests at this point:
Corpus 1 of 5: Instrumental Applicative-gen
6/6 (100.0%) tests pass (0/6 (0.0%) match gold)
Corpus 2 of 5: Locative Applicative-gen
6/6 (100.0%) tests pass (0/6 (0.0%) match gold)
Corpus 3 of 5: Factual Mood-gen
5/5 (100.0%) tests pass (3/5 (60.0%) match gold)
Corpus 4 of 5: Pronouns-gen
28/28 (100.0%) tests pass (3/28 (10.71%) match gold)
Corpus 5 of 5: Number Plural Marker-gen
1/1 (100.0%) tests pass (0/1 (0.0%) match gold)
All tests pass.
Coverage: 26.22%
Final evaluation of morphological generation
Corpus 1 of 8: Instrumental Applicative-gen
6/6 (100.0%) tests pass (5/6 (83.33%) match gold)
Corpus 2 of 8: Locative Applicative-gen
6/6 (100.0%) tests pass (2/6 (33.33%) match gold)
Corpus 3 of 8: Factual Mood-gen
5/5 (100.0%) tests pass (4/5 (80.0%) match gold)
Corpus 4 of 8: Pronouns-gen
28/28 (100.0%) tests pass (19/28 (67.86%) match gold)
Corpus 5 of 8: Number Plural Marker-gen
4/4 (100.0%) tests pass (2/4 (50.0%) match gold)
Corpus 6 of 8: Regular Verbs-gen
12/12 (100.0%) tests pass (5/12 (41.67%) match gold)
Corpus 7 of 8: Optative Mood-gen
5/5 (100.0%) tests pass (5/5 (100.0%) match gold)
Corpus 8 of 8: Spellrelax-gen
6/6 (100.0%) tests pass (0/6 (0.0%) match gold)
We added 5 twol rules.
Number of tokenised words in the corpus: 225 Coverage: 30.67% Top unknown words in the corpus: 13 ko 11 no 10 tamdaw 9 to 7 i 7 o 4 caay 4 tatiihay 3 rakat 3 sa 2 O 2 ciraraway 2 Kawas 2 Rikec 2 ka 2 mo 2 ^celay 2 Nawiro 2 kasakitakit 1 so Translation time: 0.05767536163330078 seconds
