Polish and Russian/Work plan

From Apertium
Jump to navigation Jump to search

Weekly plan

Semana Dates Coverage Testvoc Eval. (%) cov. raw (%) cov. trimmed (%) WER Bidix Err. Achieved ?
pol rus pol→rus rus→pol pol→rus rus→pol pol→rus rus→pol
1 18 abril—24 abril 76% 21,455 175,678 114,856
2 25 abril—1 mayo 78% 89.5 53.9 176620 114375
3 2 mayo—8 mayo 80% pr, cnj*, adv 500
4 9 mayo—15 mayo 80%
5 16 mayo—22 mayo 80.5%
6 23 mayo—29 mayo 81% prn, det
7 30 mayo—5 junio 81.5%
8 6 junio—12 junio 82%
9 13 junio—19 junio 83%
10 20 junio—26 junio 84%
11 27 junio—3 julio 85% n 500
12 4 julio—10 julio 86%
13 11 junio—17 julio 87% vblex
14 18 julio—24 julio 87%
13 25 junio—31 julio 88% adj
14 1 agosto—7 agosto 89%
15 8 agosto—14 agosto 90% 2000
16 15 agosto—21 agosto 90%

Calculating numbers

Errors (calculate in apertium-pol-rus)
$ sh dev/testvoc/generation.sh pol-rus | wc -l 
$ sh dev/testvoc/generation.sh rus-pol | wc -l
Bidix (calculate in apertium-pol-rus)
$ cat apertium-pol-rus.pol-rus.dix | grep '<l' | wc -l
Trimmed coverage (calculate in apertium-pol-rus)
$ cat pol.crp.txt | apertium -d . pol-rus-morph | sed 's/\$\W*\^/$\n^/g' > /tmp/pol.trim.coverage.txt
$ calc `cat /tmp/pol.trim.coverage.txt | grep -v '\*' | wc -l `/`cat /tmp/pol.trim.coverage.txt | wc -l`
Raw coverage (calculate in apertium-pol, apertium-rus)
$ cat pol.crp.txt | apertium -d . pol-morph | sed 's/\$\W*\^/$\n^/g' > /tmp/pol.raw.coverage.txt
$ calc `cat /tmp/pol.raw.coverage.txt | grep -v '\*' | wc -l `/`cat /tmp/pol.raw.coverage.txt | wc -l`

or:

$ cat pol.crp.txt | apertium -d . pol-morph | sed 's/\$\W*\^/$\n^/g' > /tmp/pol.raw.coverage.txt
$ COVERED=`cat /tmp/pol.raw.coverage.txt | grep -v '\*' | wc -l `
$ TOTAL=`cat /tmp/pol.raw.coverage.txt | wc -l`
$ echo $COVERED/$TOTAL | bc -l