Parser Self-Training for Syntax-Based Machine Translation

In syntax-based machine translation, it is known that the accuracy of parsing greatly affects the translation accuracy. Self-training, which uses parser output as training data, is one method to improve the parser accuracy. However, because parsing errors cause noisy data to be mixed with the training data, automatically generated parse trees do not always contribute to improving accuracy. In this paper, we propose a method for selecting self-training data by performing syntaxbased machine translation using a variety of parse trees, using automatic evaluation metrics to select which translation is better, and using that translation’s parse tree for parser selftraining. This method allows us to automatically choose the trees that contribute to improving translation accuracy, improving the effectiveness of self-training. In experiments, we found that our self-trained parsers significantly improve a state-of-the-art syntax-based machine translation system in two language pairs.