User:Francis Tyers/TLH

From Apertium
Jump to navigation Jump to search

Tarea 1

for x in `seq 0 9`; do 

        cooked2lex.pl < LexEsp-train-$x.cooked > train-$x.lex
        cooked2ngram.pl < LexEsp-train-$x.cooked > train-$x.ngrams
        cooked2raw.pl < LexEsp-$x.cooked > LexEsp-$x.raw
        t3 train-$x.ngrams train-$x.lex < LexEsp-$x.raw > LexEsp-$x.t3
        evaluate.pl LexEsp-$x.cooked LexEsp-$x.t3 >> output
        cat output

done

$ cat output 
418 sentences
         LexEsp-0.t3     9470      397  95.976%
418 sentences
         LexEsp-1.t3     9290      410  95.773%
418 sentences
         LexEsp-2.t3     9199      430  95.534%
418 sentences
         LexEsp-3.t3     9264      440  95.466%
418 sentences
         LexEsp-4.t3     9164      448  95.339%
418 sentences
         LexEsp-5.t3     8908      397  95.733%
418 sentences
         LexEsp-6.t3     8968      418  95.547%
418 sentences
         LexEsp-7.t3     9334      414  95.753%
418 sentences
         LexEsp-8.t3     9693      478  95.300%
417 sentences
         LexEsp-9.t3     9434      405  95.884%

Tarea 2

$ for i in `seq 1 9`; do 
    cat LexEsp-[1-$i].cooked > LexEsp-ejecucion$i.cooked; 
    cooked2lex.pl < LexEsp-ejecucion$i.cooked > train.$i.lex; 
    cooked2ngram.pl < LexEsp-ejecucion$i.cooked > train.$i.ngrams; 
    t3 train.$i.ngrams train.$i.lex < LexEsp-0.raw > LexEsp-0.$i.t3; 
    evaluate.pl LexEsp-0.cooked LexEsp-0.$i.t3 >> output ; 
done

$ wc -l LexEsp-ejecucion*.cooked
    418 LexEsp-ejecucion1.cooked
    836 LexEsp-ejecucion2.cooked
   1254 LexEsp-ejecucion3.cooked
   1672 LexEsp-ejecucion4.cooked
   2090 LexEsp-ejecucion5.cooked
   2508 LexEsp-ejecucion6.cooked
   2926 LexEsp-ejecucion7.cooked
   3344 LexEsp-ejecucion8.cooked
   3761 LexEsp-ejecucion9.cooked

$ cat output
418 sentences
       LexEsp-0.1.t3     8948      919  90.686%
418 sentences
       LexEsp-0.2.t3     9155      712  92.784%
418 sentences
       LexEsp-0.3.t3     9275      592  94.000%
418 sentences
       LexEsp-0.4.t3     9313      554  94.385%
418 sentences
       LexEsp-0.5.t3     9366      501  94.922%
418 sentences
       LexEsp-0.6.t3     9391      476  95.176%
418 sentences
       LexEsp-0.7.t3     9419      448  95.460%
418 sentences
       LexEsp-0.8.t3     9444      423  95.713%
418 sentences
       LexEsp-0.9.t3     9470      397  95.976%

Tarea 3

$ for i in `seq 1 10`; do 
    t3 -l $i train.ngrams train.lex < LexEsp-0.raw > LexEsp-0.l$i.t3; 
    evaluate.pl LexEsp-0.cooked LexEsp-0.l$i.t3 >> output.l; 
done

$ cat output.l
418 sentences
      LexEsp-0.l1.t3     9411      456  95.379%
418 sentences
      LexEsp-0.l2.t3     9466      401  95.936%
418 sentences
      LexEsp-0.l3.t3     9492      375  96.199%
418 sentences
      LexEsp-0.l4.t3     9490      377  96.179%
418 sentences
      LexEsp-0.l5.t3     9473      394  96.007%
418 sentences
      LexEsp-0.l6.t3     9477      390  96.047%
418 sentences
      LexEsp-0.l7.t3     9473      394  96.007%
418 sentences
      LexEsp-0.l8.t3     9470      397  95.976%
418 sentences
      LexEsp-0.l9.t3     9470      397  95.976%
418 sentences
     LexEsp-0.l10.t3     9470      397  95.976%

Tarea 4

$ prepare-corpus.sh LexEsp_Etq_Larga.cooked
4179 sentences
256 tags 16481 types 96961 tokens
  1     15735  95.474%     69045  71.209% 
  2       689   4.181%     22621  23.330% 
  3        51   0.309%      4315   4.450% 
  4         3   0.018%       151   0.156% 
  5         3   0.018%       829   0.855% 
Mean ambiguity A=1.361176

Entropy H(p)=5.488119

$ cooked2lex.pl < LexEsp_Etq_Larga-train-0.cooked > train.larga.lex
3761 sentences
254 tags 15431 types 87094 tokens
  1     14742  95.535%     62581  71.855% 
  2       637   4.128%     20044  23.014% 
  3        47   0.305%      3620   4.156% 
  4         2   0.013%        96   0.110% 
  5         3   0.019%       753   0.865% 
Mean ambiguity A=1.351161

Entropy H(p)=5.485330

$ cooked2ngram.pl < LexEsp_Etq_Larga-train-0.cooked > train.larga.ngrams
$ cooked2raw.pl LexEsp_Etq_Larga-0.cooked > LexEsp_Etq_Larga-0.raw
$ cooked2raw.pl < LexEsp_Etq_Larga-0.cooked > LexEsp_Etq_Larga-0.raw
$ t3 train.larga.ngrams train.larga.lex < LexEsp_Etq_Larga-0.raw > LexEsp_Etq_Larga-0.t3
[        4 ms::1] 
[        4 ms::1] Trigram POS Tagger (c) Ingo Schröder, schroeder@informatik.uni-hamburg.de
[        4 ms::1] 
[     2064 ms::1] model generated from 3761 sentences (thereof 43 one-word)
[     2064 ms::1] found 11283 uni-, 15044 bi-, and 18762 trigram counts for the boundary tag
[    12724 ms::1] computed smoothed transition probabilities
[    13512 ms::1] built suffix tries with 29924 lowercase and 6743 uppercase nodes
[    13532 ms::1] leaves/single/total LC: 7672 18878 29925
[    13536 ms::1] leaves/single/total UC: 1320 4874 6744
[    16329 ms::1] suffix probabilities smoothing done [theta 1.281e-02]
[ 12249377 ms::1] done

$ evaluate.pl LexEsp_Etq_Larga-0.cooked LexEsp_Etq_Larga-0.t3

418 sentences
LexEsp_Etq_Larga-0.t3     9412      455  95.389%