what aspects of language get considered while evaluating the translation by COMET. as far as I am aware of MQM metrics are covered in it. Interested in knowing other terminology/features of language gets in COMET. I am looking for metric along with COMET that can assess the translated content for cosmetic domain data. please suggest. #175
I am not sure I understand the question very well. Can you give a specific example of what you want to evaluate?
COMET models are trained with different signals. Most models are trained with DA which basically measure adequacy of the output. Other models are trained with MQM. Models trained with MQM should ideally model all dimensions covered by the MQM typology.
I am not sure I understand the question very well. Can you give a specific example of what you want to evaluate?
COMET models are trained with different signals. Most models are trained with DA which basically measure adequacy of the output. Other models are trained with MQM. Models trained with MQM should ideally model all dimensions covered by the MQM typology.