Open iamkittykat opened 4 months ago
you're right, it tried it out too,i got the same outcome
@iamkittykat @joschan21 , i think it is because the model splits the input into 3 characters then test the characters again known profane words and it ignores white spaces, for example "its h" would get split into something like this [ its, ith, ist, ish, iht, ihs, tis, tih, tsi, tsh, thi, ths, sit, sih, sti, sth, shi, sht, hit, his, hti, hts, hsi, hst ] the it will test all the items in the array against permutations of known profane words in this case "shit" and when it sees most of the items in the array match those of "shit" it marks it as profane
i would have fixed it but the dataset for the model are too large for me to download
okay
like "its h" gets for sh_t same for n and more!