Difference between revisions of "User:Firespeaker/GSoC2014/Progress"

From Apertium
Jump to navigation Jump to search
(→‎WER: r52183)
Line 173: Line 173:
 
!colspan="4"|texts
 
!colspan="4"|texts
 
! week
 
! week
! 0a !! 0b !! 1 !! 2
+
! 0a !! 0b !! 0c !! 1 !! 2
 
|-
 
|-
 
!rowspan="2"| name
 
!rowspan="2"| name
Line 180: Line 180:
 
!rowspan="2"| direction
 
!rowspan="2"| direction
 
! date
 
! date
| 2014-03-21 || 2014-04-06 || 2014-××-×× || 2014-××-××
+
| 2014-03-21 || 2014-04-06 || 2014-04-21 || 2014-××-×× || 2014-××-××
 
|-
 
|-
 
! revision
 
! revision
| r51024 || r51765 || r××××× || r×××××
+
| r51024 || r51765 || r52183 || r××××× || r×××××
 
|-
 
|-
 
| foo || kaz || ~200 || kaz-kir ||rowspan="6" align="center"|dev set 1
 
| foo || kaz || ~200 || kaz-kir ||rowspan="6" align="center"|dev set 1
Line 200: Line 200:
 
| 99.46% ~ 98.66%
 
| 99.46% ~ 98.66%
 
| 58.06% ~ 49.19%
 
| 58.06% ~ 49.19%
  +
| 66.67% ~ 54.85%
 
|-
 
|-
 
| tur-uzb
 
| tur-uzb

Revision as of 03:55, 5 June 2014

trimmed coverage

week 1 2
date 2014-××-×× 2014-××-××
revision r××××× r×××××
kaz(-kir)
kir(-kaz)
tur(-kir)
kir(-tur)
tur(-uzb)
uzb(-tur)

monodix stems

week 0a 0b 0c 1 2
date 2014-03-21 2014-04-06 2014-04-21 2014-××-×× 2014-××-××
revision r51024 r51765 r52183 r××××× r×××××
kaz 11332 11336 11337
kir 13637 13703 13705
tur 11128 11186 11172
uzb 3922 3957 3957

bidix stems

week 0a 0b 0c 1 2
date 2014-03-21 2014-04-06 2014-04-21 2014-××-×× 2014-××-××
revision r51024 r51765 r52183 r××××× r×××××
kaz-kir 7557 7557 7557
tur-kir 7163 7249 7107
tur-uzb 2416 2416 2416

CG per-token ambiguity

tokens( analyser | CG ) / tokens( analyser )

week 0a 0b 0c 1 2
date 2014-03-21 2014-04-06 2014-04-21 2014-××-×× 2014-××-××
revision r51024 r51765 r52183 r××××× r×××××
kaz
kir
tur SETimes 1.76 → 1.5 2.01 → 1.28 2.32 → 1.29
uzb

lrx per-token ambiguity

tokens( analyser | CG | biltrans | lrx ) / tokens( analyser | CG | biltrans )

week 0a 0b 1 2
date 2014-03-21 2014-04-06 2014-××-×× 2014-××-××
revision r51024 r51765 r××××× r×××××
kaz(-kir)
kir(-kaz)
tur(-kir) SETimes 1.13155 → 1.01861 1.1099 → 1.02028
kir(-tur)
tur(-uzb)
uzb(-tur)

corpus testvoc

week 0a 0b 1 2
date 2014-03-21 2014-04-06 2014-××-×× 2014-××-××
revision r51024 r51765 r××××× r×××××
kaz(-kir)
kir(-kaz)
tur(-kir) SETimes 19.67% 6.72%
kir(-tur)
tur(-uzb)
uzb(-tur)

WER

texts week 0a 0b 0c 1 2
name language № words direction date 2014-03-21 2014-04-06 2014-04-21 2014-××-×× 2014-××-××
revision r51024 r51765 r52183 r××××× r×××××
foo kaz ~200 kaz-kir dev set 1
foo kir ~200 kir-kaz
kir-tur
küçükkuş tur 339 tur-kir 99.46% ~ 98.66% 58.06% ~ 49.19% 66.67% ~ 54.85%
tur-uzb
foo uzb ~200 uzb-tur
bar kaz ~200 kaz-kir dev set 2
bar kir ~200 kir-kaz
kir-tur
bar tur ~200 tur-kir
tur-uzb
bar uzb ~200 uzb-tur
baz kaz ~500 kaz-kir dev set 3
baz kir ~500 kir-kaz
kir-tur
baz tur ~500 tur-kir
tur-uzb
baz uzb ~500 uzb-tur
foo kaz ~200 kaz-kir eval set 1
foo kir ~200 kir-kaz
kir-tur
foo tur ~200 tur-kir
tur-uzb
foo uzb ~200 uzb-tur
bar kaz ~200 kaz-kir eval set 2
bar kir ~200 kir-kaz
kir-tur
bar tur ~200 tur-kir
tur-uzb
bar uzb ~200 uzb-tur
baz kaz ~500 kaz-kir eval set 3
baz kir ~500 kir-kaz
kir-tur
baz tur ~500 tur-kir
tur-uzb
baz uzb ~500 uzb-tur