Improvement Research on Computer Assistant Proverb Translation

摘要

The thesis focuses on the efficiency problem caused by the unordered reference version provided by CAT software when calling the translation memory database. The BP neural network is used to study the connection between the translator's background and his/her preference for the goal of providing translators with ordered reference version. The thesis first acquires the research datum of the translator's background and their preference of translation. And the datum are quantified by options. Then, based on the BP neural network, an analysis model telling the preference between foreignizing and domesticating translation will be set up. Furthermore, the neural network will be trained through the training data set. Finally, with the help of detected set, we can check the grading accuracy between the predicted results and the expected value. Analysis model of the preference between foreignizing translation and domesticating translation of translators based on BP neural network can better recognize the translator's foreignizing and domesticating translation preference.