id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_fwtbvrrkgzhozg6jd22ortpemu	M. R. Costa-jussa	Towards human linguistic machine translation evaluation	2013.0	14	.pdf	application/pdf	4916	571	64	When evaluating machine translation outputs, linguistics is usually taken into Evaluation of a translation output is not an easy task even for human beings because translation involves different types of knowledge, such as linguistic machine translation evaluation metric for DARPA's Global Autonomous Language Exploitation (GALE) program [12] was HTER (Human-targeted Translation Edit Rate). correlation results at the segment level between linguistic evaluations and different types of systems. The linguistic guidelines have been designed for the Catalan target language This error is detected when there is no correspondence at all between the source word and the translated target word. 5. Missing target word This refers to a non-translated source word, Finally, we compute the correlation of the linguistic evaluations: among different translation systems, and with standard human Table 4: Linguistic evaluation results for English-to-Catalan translation outputs: number and In this sense, linguistic guidelines have been shown useful for machine translation evaluation.	./cache/work_fwtbvrrkgzhozg6jd22ortpemu.pdf	./txt/work_fwtbvrrkgzhozg6jd22ortpemu.txt