Difference between revisions of "Talk:Scandinavian MT project"

From Apertium
Jump to navigation Jump to search
 
Line 7: Line 7:
 
nno-swe
 
nno-swe
 
unk known tot cov %
 
unk known tot cov %
876166 5251500 6127666 85.7015 no decomp ex-upper: 87.6831
+
805040 5307135 6112175 86.8289 no decomp ex-upper: 88.251
773219 5354447 6127666 87.3815 with decomp ex-upper: 91.0789
+
703448 5408727 6112175 88.491 with decomp ex-upper: 91.5592
   
 
nob-swe
 
nob-swe
 
unk known tot cov %
 
unk known tot cov %
744696 5369943 6114639 87.8211 no decomp ex-upper: 89.4771
+
691359 5408402 6099761 88.6658 no decomp ex-upper: 90.0757
632117 5482522 6114639 89.6622 with decomp ex-upper: 92.8601
+
579262 5520499 6099761 90.5035 with decomp ex-upper: 93.3714
   
 
swe-nno
 
swe-nno
 
unk known tot cov %
 
unk known tot cov %
992577 5282677 6275254 84.1827 no decomp ex-upper: 87.128
+
914584 5167659 6082243 84.963 no decomp ex-upper: 87.082
813461 5461794 6275255 87.037 with decomp ex-upper: 92.1041
+
744040 5338204 6082244 87.767 with decomp ex-upper: 92.1901
   
 
swe-nob
 
swe-nob
 
unk known tot cov %
 
unk known tot cov %
979754 5295424 6275178 84.3868 no decomp ex-upper: 87.3452
+
907104 5175868 6082972 85.0878 no decomp ex-upper: 87.2664
795866 5479313 6275179 87.3172 with decomp ex-upper: 92.4016
+
733559 5349414 6082973 87.9408 with decomp ex-upper: 92.4382
   
 
dan-swe
 
dan-swe
 
unk known tot cov %
 
unk known tot cov %
1168922 4733019 5901941 80.1943 with decomp ex-upper: 87.5534
+
1135679 4775050 5910729 80.7861 no decomp ex-upper: 86.2313
1301693 4600248 5901941 77.9447 no decomp ex-upper: 82.9197
+
997339 4913390 5910729 83.1266 with decomp ex-upper: 90.5919
   
 
swe-dan
 
swe-dan
 
unk known tot cov %
 
unk known tot cov %
1337817 4539056 5876873 77.2359 no decomp ex-upper: 81.2688
+
1120955 4760872 5881827 80.9421 no decomp ex-upper: 86.4166
1136015 4740858 5876873 80.6697 with decomp ex-upper: 88.3335
+
926361 4955466 5881827 84.2505 with decomp ex-upper: 92.2963
   
 
dan-nno
 
dan-nno
 
unk known tot cov %
 
unk known tot cov %
869064 5033026 5902090 85.2753 no decomp ex-upper: 87.573
+
870318 5020221 5890539 85.2252 no decomp ex-upper: 87.512
750810 5151280 5902090 87.2789 with decomp ex-upper: 91.5197
+
752064 5138475 5890539 87.2327 with decomp ex-upper: 91.4789
   
 
dan-nob
 
dan-nob
 
unk known tot cov %
 
unk known tot cov %
831676 5065778 5897454 85.8977 no decomp ex-upper: 88.2809
+
832984 5052919 5885903 85.8478 no decomp ex-upper: 88.22
702310 5195144 5897454 88.0913 with decomp ex-upper: 92.3478
+
703618 5182285 5885903 88.0457 with decomp ex-upper: 92.308
   
 
nno-dan
 
nno-dan
 
unk known tot cov %
 
unk known tot cov %
708015 5304680 6012695 88.2247 no decomp ex-upper: 90.2486
+
707162 5295158 6002320 88.2185 no decomp ex-upper: 90.2369
614142 5398553 6012695 89.7859 with decomp ex-upper: 93.4224
+
613734 5388586 6002320 89.7751 with decomp ex-upper: 93.4124
   
 
nob-dan
 
nob-dan
 
unk known tot cov %
 
unk known tot cov %
594853 5397002 5991855 90.0723 no decomp ex-upper: 91.7318
+
594784 5387514 5982298 90.0576 no decomp ex-upper: 91.7191
492242 5499613 5991855 91.7848 with decomp ex-upper: 94.8782
+
492014 5490284 5982298 91.7755 with decomp ex-upper: 94.8719
   
 
nno-nob
 
nno-nob
 
unk known tot cov %
 
unk known tot cov %
706329 5367221 6073550 88.3704 no decomp ex-upper: 91.2201
+
704975 5364593 6069568 88.3851 no decomp ex-upper: 91.2337
614665 5458885 6073550 89.8796 with decomp ex-upper: 94.2577
+
613334 5456234 6069568 89.8949 with decomp ex-upper: 94.2736
   
 
nob-nno
 
nob-nno
 
unk known tot cov %
 
unk known tot cov %
603978 5459984 6063962 90.0399 no decomp ex-upper: 92.6016
+
603139 5456170 6059309 90.0461 no decomp ex-upper: 92.6105
503705 5560257 6063962 91.6935 with decomp ex-upper: 95.5487
+
502464 5556845 6059309 91.7076 with decomp ex-upper: 95.5668
 
</pre>
 
</pre>
  +
 
===script===
 
===script===
 
<pre>
 
<pre>

Latest revision as of 18:20, 1 July 2016

Coverage[edit]

Coverage on Wikipedia dumps ("no decomp" is with decompounding turned off, ie. without the -e switch to lt-proc; ex-upper is excluding anything with uppercase characters from all counts; script at bottom of page).

Results[edit]

$ ./scandicov.sh
nno-swe
unk     known   tot     cov %
805040  5307135 6112175 86.8289   no decomp     ex-upper:       88.251
703448  5408727 6112175 88.491  with decomp     ex-upper:       91.5592

nob-swe
unk     known   tot     cov %
691359  5408402 6099761 88.6658   no decomp     ex-upper:       90.0757
579262  5520499 6099761 90.5035 with decomp     ex-upper:       93.3714

swe-nno
unk     known   tot     cov %
914584  5167659 6082243 84.963    no decomp     ex-upper:       87.082
744040  5338204 6082244 87.767  with decomp     ex-upper:       92.1901

swe-nob
unk     known   tot     cov %
907104  5175868 6082972 85.0878   no decomp     ex-upper:       87.2664
733559  5349414 6082973 87.9408 with decomp     ex-upper:       92.4382

dan-swe
unk     known   tot     cov %
1135679 4775050 5910729 80.7861   no decomp     ex-upper:       86.2313
997339  4913390 5910729 83.1266 with decomp     ex-upper:       90.5919

swe-dan
unk     known   tot     cov %
1120955 4760872 5881827 80.9421   no decomp     ex-upper:       86.4166
926361  4955466 5881827 84.2505 with decomp     ex-upper:       92.2963

dan-nno
unk     known   tot     cov %
870318  5020221 5890539 85.2252   no decomp     ex-upper:       87.512
752064  5138475 5890539 87.2327 with decomp     ex-upper:       91.4789

dan-nob
unk     known   tot     cov %
832984  5052919 5885903 85.8478   no decomp     ex-upper:       88.22
703618  5182285 5885903 88.0457 with decomp     ex-upper:       92.308

nno-dan
unk     known   tot     cov %
707162  5295158 6002320 88.2185   no decomp     ex-upper:       90.2369
613734  5388586 6002320 89.7751 with decomp     ex-upper:       93.4124

nob-dan
unk     known   tot     cov %
594784  5387514 5982298 90.0576   no decomp     ex-upper:       91.7191
492014  5490284 5982298 91.7755 with decomp     ex-upper:       94.8719

nno-nob
unk     known   tot     cov %
704975  5364593 6069568 88.3851   no decomp     ex-upper:       91.2337
613334  5456234 6069568 89.8949 with decomp     ex-upper:       94.2736

nob-nno
unk     known   tot     cov %
603139  5456170 6059309 90.0461   no decomp     ex-upper:       92.6105
502464  5556845 6059309 91.7076 with decomp     ex-upper:       95.5668

script[edit]

$ cat scandicov.sh
#!/bin/bash

# kill process group:
trap "kill -- -0" EXIT

sum () {
    awk -v note="$1" '
BEGIN{OFS=FS="\t"} 
/^\^/{w++} 
/\/\*/{u++}
/\/\*.*[[:upper:]]/{Uu++}
/[[:upper:]]/{Uw++} 
END{
lw=w-Uw
lu=u-Uu
print u,w-u,w,100*(w-u)/w, note,"ex-upper:",100*(lw-lu)/lw}'
}

for ana in /l/a/*/apertium-{swe-nor,swe-dan,dan-nor,nno-nob}/*.automorf.bin; do
    f="$(basename "${ana}")"
    echo "${f%%.automorf.bin}"
    src="${f%%-*}"
    printf "unk\tknown\ttot\tcov %%\n"
    xzcat ~/corpora/"${src}".wikicov.xz | lt-proc "${ana}" \
        | apertium-cleanstream -n \
        | sum "  no decomp" &

    xzcat ~/corpora/"${src}".wikicov.xz | lt-proc -e "${ana}" \
        | apertium-cleanstream -n \
        | sum "with decomp" &

    wait
    echo
done

Testvoc[edit]

Results[edit]

   1 ../apertium-dan-nor/nobackup/69419-dan-nno
  30 ../apertium-dan-nor/nobackup/69419-dan-nob
  37 ../apertium-dan-nor/nobackup/69419-nno-dan
  26 ../apertium-dan-nor/nobackup/69419-nob-dan
  94 total

  136 ../apertium-nno-nob/nobackup/69419-nno-nob
   71 ../apertium-nno-nob/nobackup/69419-nob-nno
  207 total

  157 ../apertium-swe-dan/nobackup/69419-dan-swe
  276 ../apertium-swe-dan/nobackup/69419-swe-dan
  433 total

0 ../apertium-swe-nor/nobackup/69419-nno-swe
0 ../apertium-swe-nor/nobackup/69419-nob-swe
0 ../apertium-swe-nor/nobackup/69419-swe-nno
0 ../apertium-swe-nor/nobackup/69419-swe-nob.tmp
0 total

script[edit]

$ while sleep 30; do 
  for d in trunk/apertium-{swe-nor,dan-nor,nno-nob,swe-dan}; do 
    ( cd $d && up && make langs  && rev=$(svn info|grep ^Revision: |grep -o '[0-9]*') && 
      for m in modes/???-???.mode; do 
        p=${m##modes/} && p=${p%%.mode} && echo $p && 
        ( test -f nobackup/${rev}-$p || 
            ( dev/testvoc/generation.sh --hfst $p > nobackup/${rev}-$p.tmp
              mv nobackup/${rev}-$p.tmp nobackup/${rev}-$p 
            ) 
        )
      done  
    )
  done
done