Institute of Formal and Applied Linguistics
Charles University, Czech Republic
Faculty of Mathematics and Physics
Utilization of coreference in MT
Utilization of coreference in Machine Translation
Majority of present machine translation systems do not address the retaining of text coherency, they translate just isolated sentences. On the other hand, the authors of anaphora resolvers rarely integrate these tools into more complex systems, e.g. the machine translation systems. The proposed project focuses on the research of ways, how to incorporate automatic resolution of coreferential and other anaphoric relations into machine translation in order to improve the quality of translation. The proposers will try to exploit the knowledge about anaphoric relations on the side of the source language as well as on the side of the target one. The methods of supervised machine learning will be utilized to model anaphoric relations. Due to easy access to unannotated data the research will be concerned with the use of unsupervised learning methods. The success rate of the proposed methods will be experimentally tested on the English-Czech language pair using the standard measures for machine translation evaluation.