D 4 . 1 b MT systems for the second development cycle

Executive Summary This deliverable D4.1b describes work done in the area of machine translation development and deployment (work package 4) during the period of the second milestone of the project (from month 7 to month 24). As part of this second development cycle, we have built a new generic MT system by building upon the system we prepared for milestone 1; we have then broadened our scope by (i) building a MT system for a specific domain (tourism) and (ii) experimenting with the application of linguistic knowledge, through a process called morph segmentation, to avoid the data sparsity caused by unknown word forms which can however be morphologically broken into known tokens; we have developed a comprehensive architecture to provide these machine translation systems through a web interface. The deliverable also briefly reports on other related activities in this work package, namely (i) our participation in the WMT14 translation contest and (ii) experiments with a procedure to clean noisy parallel corpora.