============================================================================================
EVALUATION module (dj.py) .... NOTE: execution may take 45-50seconds for single magazine
First argument is manually aligned texts
Second argument is automatically aligned text (HUN or GARG) text
Third argument is similarity ration (0..100) used by FUZZYWUZZY module. 100 means identical.
Result is very verbose, and includes RECALL and PRECISION at the end.
python ./dj.py ./Mng56_en_sr_manual_dj.txt ./Mng56_garg_fix.txt 95 > report_garg.txt
tail -n 10 ./report_garg.txt
============================================================================================
Following command will report all sentences present in manual src list, but not present in eval (hun or garg) src list:
grep "MANUAL_NOTFOUND" ./report_garg.txt

Following command will report all sentences present in eval (hun or garg) src list, but not present in manual src list:
grep "EVAL_EXTRA" ./report_garg.txt

Following command will report all bad precision cases (two BAD_LINE_M indicates wrong manual line (src and tgt) and following BAD_LINE_E indicates wrong eval line
grep "BAD_LINE_" ./report_garg.txt

============================================================================================
Another module prepares manually aligned texts:
./tmx/between_tag_extractor.py OR
./tmx/dana_between_tag_extractor.py
... which converts TMX files to txt files with sentences only.


