As machine translation (MT) systems grow more complex and incorporate more linguistic knowledge, ������� � it becomes more difficult to evaluate independent pieces of the MT pipeline. Being able� � to������� � inspect�many ������� � of ������ � the intermediate �� � ������������ � data structures ������ � ����������� � used during ��MT ������ � decoding � � ������ � allows �� a��� � more ���fine-grained ��������� � �� � evaluation �� � �������� � of MT ��������� � performance, ���������helping � � ������� � to determine ���� � �� � which �� � ���parts ��������� � of the current ������ � ������������� � process are effective n���� � ������� � and which ���� � are����������� � not. In this ���� � article, �� � we ������������ � present an�� � overview k���� � ������ � of the visualization ���� � ������������ � tools������� � that are ���currently ���������� � distributed ���������with ��������� � the Joshua �� � �������� � (Li et al., � � 2009) ��� � � � MT �������� � decoder. �� We ���������� � explain ������������� � their use and��� � present �� � ������ � an example � � ����of ��� � how � � visually ���� � ����� � inspecting ��� � � ������� � the decoder’s ������ � ������ � data

We developed caitra, a novel tool that aids human translators by (a) making suggestions for sentence completion in an interactive machine translation setting, (b) providing alternative word and phrase translations, and (c) allowing them to postedit machine translation output. The tool uses the Moses decoder, is implemented in Ruby on Rails and C++ and delivered over the web. 1

We carried out a study on monolingual translators with no knowledge of the source language, but aided by post-editing and the display of translation options. On Arabic-English and Chinese-English, using standard test data and current statistical machine translation systems, 10 monolingual translators were able to translate 35 % of Arabic and 28 % of Chinese sentences correctly on average, with some of the participants coming close to professional bilingual performance on some of the documents. While machine translation systems have advanced greatly over the last decade, nobody seriously expects human-level performance any time soon, except for very constraint settings. But are todays systems good enough to enable monolingual speakers of the target language without knowledge of the source language to generate correct translations? And what type of assistance from machine translation is most helpful for such translators? We carried out a study that involved monolingual translators who had no knowledge of Chinese and Arabic to translate documents from the NIST 20081 test sets, being assisted by statistical machine translation systems trained on data created under the GALE2 research program. Our study shows that monolingual translators were able to translate 35 % of Arabic and 28 % of Chinese sentences, under a strict standard of correctness that scored professional bilingual translations as 61 % and 66 % correct for Arabic and Chinese, respectively. We found also large variability among the participants and between the documents in the

We introduce a new approach to the problem of obtaining cost-effective, reasonable quality translation, by exploiting simple and inexpensive human computations by monolingual speakers. The key insight behind the process is that it is possible to to spot likely translation errors with only monolingual knowledge of the target language, and it is possible to generate new ways to say the same thing (i.e. paraphrases) with only monolingual knowledge of the source language. Initial evaluation demonstrates substantial improvements in translation quality. 1.

Abstract—Machine Translation (MT) systems are evaluated and debugged using the BLEU automated metric. However, the current community implementation of BLEU is not ideal for MT system developers and researchers since it only produces textual information. I present a novel tool called iBLEU that organizes BLEU scoring information in a visual and easy-tounderstand manner, making it easier for MT system developers & researchers to quickly locate documents and sentences on which their system performs poorly. It also allows comparing translations from two different MT systems. Furthermore, one can also choose to compare to the publicly available MT systems, e.g., Google Translate and Bing Translator, with a single click. It can run on all major platforms and requires no setup whatsoever. I.

Abstract. We investigate novel types of assistance for human translators, based on statistical machine translation methods. We developed the computer aided tool Caitra that makes suggestions for sentence completion, shows word and phrase translation options, and allows postediting of machine translation output. We carried out a study of the translation process that involved non-professional translators that were native in either French or English and recorded their interaction with the tool. Users translated 192 sentences from French news stories into English. Most translators were faster and better when using assistance. A detailed examination of the logs also provides insight into the human translation process, such as time spent on different activities and length of pauses.

proportions, the flags adopted in 1889 are still in schizophrenia, the condition manifested itself before use by the imperial family. A person diagnosed with schizophrenia may The Efficacy of Human Post-Editing for Language Translation experience hallucinations (most reported are hearing voices), delusions (often bizarre or