PLTranslationEmpirical
PLTranslationEmpirical copied to clipboard
Query about the cleaning code
Hey~, I got some problems with the cleaning of generation codes. It would be appreciated if you could help me out. After I used clean_generations.py code to clean the generated translations of StarCoder, I found the quality of cleaning very poor and the codes can not achieve the results in the Artifacts RQ1 . Are there any other post-processing techniques to use before testing the translation performances?
Can you please give more details? Did you use the cleaning script on our starcoder generated translations? Have you checked the starcoder translations from artifacts website on zenodo?