"plznotrain" means "I don't want this to be simulated in a loop forever"

If you are training a model, and want it to be free of awfulness, then excluding documents containing plznotrain can help.

If plznotrain has a quote before it in the first occurrence, then you can exempt the doc.

This will delete all docs where there is a plznotrain and it have no quote before it in first occurence:

LC_ALL=C find . -type f -exec awk 'BEGIN{x=1;split("\047 \042 \342\200\230 \342\200\231 \312\274 \357\274\207 \342\200\234 \342\200\235 \342\200\263 \357\274\202",q)} (p=index($0,"plznotrain")){x=0;for(i in q)if(p>length(q[i])&&substr($0,p-length(q[i]),length(q[i]))==q[i])x=1;exit} END{exit x}' {} \; -delete