run_eval_cohesion.sh

#!/bin/bash


dir_path=$1

lang=$2
src_lang=${lang%%-*}
tgt_lang=${lang##*-}

data_path=/data/wyt/codes/DocDPO/data/2017-01-ted-test/$lang

for i in {0..11}; do
    # source=$data_path/test.en.$i
    target=$dir_path/IWSLT17.TED.tst2017.${lang}.${src_lang}.$i.${tgt_lang}
    reference=$data_path/IWSLT17.TED.tst2017.${lang}.${tgt_lang}.$i
    result=$dir_path/cohesion.txt

    # echo $target

    python -u /data/wyt/codes/DocDPO/evaluator/fine_grained_multi_demensional/eval_cohesion.py \
        --model gpt-4.1 \
        --input_file $target \
        --reference_file $reference \
        --target_language $tgt_lang \
        --output_file $result
done

python /data/wyt/codes/DocDPO/evaluator/fine_grained_multi_demensional/calc_avg_cohesion.py $dir_path/cohesion.txt

# cd $dir_path/aligned
# file_nums=$(ls test.*-s | sort -n -t . -k 2 | xargs wc -l | head -n -1 | awk '{ print $1 }')
# echo $file_nums

# cat $(ls test.*-t | sort -n -t . -k 2) > $dir_path/whole.hyp
# cat $(ls test.*-s | sort -n -t . -k 2) > $dir_path/whole.src

# sh /data/wyt/codes/DocMTAgent/consistency_evaluation/run_eval_sep.sh \
#     $dir_path/whole.src $dir_path/whole.hyp $dir_path/consistency $lang "$file_nums"

# cd $work_dir
# python indiversity_sep.py -r $dir_path/whole.src.record_sep.json | tee -a $dir_path/whole.src.consistency_sep