The paper introduces a new evaluation metric for machine translation (MT) called Translation Edit Rate (TER), which measures the number of edits required to transform a system output into a reference translation. The authors propose a human-targeted variant of TER (HTER) that uses human annotations to create targeted references, aiming to improve the correlation between the metric and human judgments of MT quality. HTER is shown to correlate highly with human judgments, outperforming other metrics like BLEU and METEOR. The study also demonstrates that HTER reduces the edit rate by 33% compared to using four untargeted references. The results suggest that HTER is a valuable tool for evaluating MT systems, providing a more intuitive and accurate measure of translation quality compared to traditional methods.The paper introduces a new evaluation metric for machine translation (MT) called Translation Edit Rate (TER), which measures the number of edits required to transform a system output into a reference translation. The authors propose a human-targeted variant of TER (HTER) that uses human annotations to create targeted references, aiming to improve the correlation between the metric and human judgments of MT quality. HTER is shown to correlate highly with human judgments, outperforming other metrics like BLEU and METEOR. The study also demonstrates that HTER reduces the edit rate by 33% compared to using four untargeted references. The results suggest that HTER is a valuable tool for evaluating MT systems, providing a more intuitive and accurate measure of translation quality compared to traditional methods.