Task 3 - calculate wer per line

This commit is contained in:
s452111 2020-04-22 13:44:10 +02:00
parent 7f9878ec18
commit 449da50f92

View File

@ -12,8 +12,11 @@ awk 'BEGIN{FS=OFS="\t"}{print $0,"(sp1_"NR")"}' < hypothesis.txt > hypothesis.tr
# use sclite to calculate WER, actually we need only lines starting with 'Scores' # use sclite to calculate WER, actually we need only lines starting with 'Scores'
sclite -f 0 -r reference.trn trn -h hypothesis.trn trn -e utf-8 -i rm -o all stdout | grep "Scores:" > wer_results.txt sclite -f 0 -r reference.trn trn -h hypothesis.trn trn -e utf-8 -i rm -o all stdout | grep "Scores:" > wer_results.txt
cat wer_results.txt | awk '{print ( ($7 + $8 + $9) / ($7 + $8 + &6) ) * 100;}' >> wer_per_line.txt cat wer_results.txt | awk '{print ( ($7 + $8 + $9) / ($7 + $8 + $6) ) * 100;}' >> wer_per_line.txt
paste wer_per_line.txt wikiniews_results.tsv paste wer_per_line.txt wikiniews_results.tsv
# calculate mean WER for all records # calculate mean WER for all records
awk '{sum += &1; n++} END { print sum / n; }' < wer_per_lines.txt > wer.txt awk '{sum += $1; n++} END { print sum / n; }' < wer_per_lines.txt > wer.txt
# calculate SSR