Browse by topic
Type of publication
Interactive Post-Editing in Machine Translation. Master’s Degree in Artificial Intelligence, Pattern Recognition and Digital Imaging. Universitat Politècnica de València. 2015. Advisor: Francisco CasacubertaThe current state of the art in Machine Translation (MT) is far from being good enough, with a post-process carried out by a human agent being necessary in many cases in order to correct translations. Statistical post-editing of a MT system has been used in the past to improve the translation quality of that system. Additionally, research on interactive translation prediction has been done with the aim of reducing the human post-editing effort. In this thesis, a new methodology that combines both techniques is proposed in order to, given a MT system, increase the translation quality of that system and reduce the effort that the human agent needs to make in order to correct the translation of that system. This methodology is tested on different scenarios (to connect with the output of a rule-based machine translation system, and as a method to adapt an statistical MT system from one domain to another) with different corpora, obtaining very encouraging results.