Research On Automatic Evaluation Technology Of English Translation Quality

The flexibility and complexity of one language result in the difficulty of quality evaluation of Machine Translation. As an earlier methodology of translation quality evaluation, manual proofreading can obtain a high checking accuracy. While, huge financial cost and much more time are required. And the checking varies with translators differing and time going by, which causes the checking process less objective. What is worse, manual proofreading cannot be reused. For all the above, the automatic evaluation technology of translation quality comes into being under the requirement.Currently, the existing MTE theories are mainly used to compare the performance of different systems by computing the similarity between the candidate translation and the reference. However, because of their dependency on the references, they are restrained for application and cannot be transformed into commercial value. With a view to this, we do some research on automatic translation quality evaluation without references.The research works of this thesis mainly include three aspects as follows: An evaluation method of English subject-verb agreement based on dependency grammar, which discovers the subject and the predicate verb by doing phrase parsing and sentence simplification operation and evaluates their agreement according to the Part of Speeches (POSs) of the subject-verb; Checking of verb forms misuse based on Conditional Random Fields (CRF) and Transform-based Error-driven, with the re-tagging of POSs following the operation of word stemming and before the post-correction of inappropriately generated verb POSs; An automatic English translation quality evaluation system with modification and supplement to the SVM features that improves the system performance, which is the advanced version of the existing evaluation system in the lab.The research results above can be applied into the evaluation of computer-assisted translation and machine translation by discovering the grammar errors that appear in the translation. In addition, for its running without references, the works have a bright application future.