Table 4

Overview of evaluation methods for each criterion.

CriterionAutomatic EvaluationHuman Evaluation
Overall BLEU with gold references Rating or ranking 
 - Transferred Style Strength Accuracy by a separately trained style classifier Rating or ranking 
 - Semantic Preservation BLEU/ROUGE/etc. with (modified) inputs Rating or ranking 
 - Fluency Perplexity by a separately trained language model Rating or ranking 
CriterionAutomatic EvaluationHuman Evaluation
Overall BLEU with gold references Rating or ranking 
 - Transferred Style Strength Accuracy by a separately trained style classifier Rating or ranking 
 - Semantic Preservation BLEU/ROUGE/etc. with (modified) inputs Rating or ranking 
 - Fluency Perplexity by a separately trained language model Rating or ranking 
Close Modal

or Create an Account

Close Modal
Close Modal