measure(ment),2-1-E06-1031,bq context </term> . Most state-of-the-art <term> evaluation measures </term> for <term> machine translation </term>
tech,5-1-E06-1031,bq <term> evaluation measures </term> for <term> machine translation </term> assign high <term> costs </term> to movements
other,9-1-E06-1031,bq machine translation </term> assign high <term> costs </term> to movements of <term> word </term> blocks
other,13-1-E06-1031,bq high <term> costs </term> to movements of <term> word </term> blocks . In many cases though such
other,13-2-E06-1031,bq result in correct or almost correct <term> sentences </term> . In this paper , we will present
measure(ment),9-3-E06-1031,bq this paper , we will present a new <term> evaluation measure </term> which explicitly models <term> block
tech,14-3-E06-1031,bq measure </term> which explicitly models <term> block reordering </term> as an <term> edit operation </term> .
tech,18-3-E06-1031,bq <term> block reordering </term> as an <term> edit operation </term> . Our <term> measure </term> can be exactly
measure(ment),1-4-E06-1031,bq an <term> edit operation </term> . Our <term> measure </term> can be exactly calculated in <term>
other,7-4-E06-1031,bq </term> can be exactly calculated in <term> quadratic time </term> . Furthermore , we will show how
measure(ment),7-5-E06-1031,bq Furthermore , we will show how some <term> evaluation measures </term> can be improved by the introduction
other,16-5-E06-1031,bq be improved by the introduction of <term> word-dependent substitution costs </term> . The correlation of the new <term>
measure(ment),5-6-E06-1031,bq </term> . The correlation of the new <term> measure </term> with <term> human judgment </term> has
other,7-6-E06-1031,bq of the new <term> measure </term> with <term> human judgment </term> has been investigated systematically
other,16-6-E06-1031,bq investigated systematically on two different <term> language pairs </term> . The experimental results will show
other,12-7-E06-1031,bq outperforms state-of-the-art approaches in <term> sentence-level correlation </term> . Results from experiments with <term>
other,4-8-E06-1031,bq </term> . Results from experiments with <term> word dependent substitution costs </term> will demonstrate an additional increase
measure(ment),16-8-E06-1031,bq additional increase of correlation between <term> automatic evaluation measures </term> and <term> human judgment </term> . In
other,20-8-E06-1031,bq automatic evaluation measures </term> and <term> human judgment </term> . In this paper , we investigate
skrij podrobnosti