search

google-research-datasets / clang8

cLang-8 is a dataset for grammatical error correction.
100 stars 5 forks source link

The task prefix of the grammatical error correction task #7

Open ryh95 opened 2 years ago

ryh95 commented 2 years ago

Dear authors of clang-8,

I want to use clang-8 to fine tune the T5 model, however, I dont know what is the task prefix (e.g. for translation, the prefix could be translate English to Romanian: )

Would you share it? Thank you very much!

ekQ commented 2 years ago

Hi, for English-only experiments (Table 4), we didn't use any prefix. For the multilingual experiments (Table 2), I think we used gec: prefix, but I haven't double checked. However, I think that the prefix doesn't really matter when you fine-tune on a single task only.