Yes-No Answers, Partial Pro-drop Languages and Machine Translation

This paper addresses the automatic translations of verbal answers to yes-no questions from partial pro-drop languages (Brazilian Portuguese and Russian) into a non-pro-drop language (English). The outputs provided by standard statistical machine translations are mostly grammatically inaccurate or semantic-pragmatically inadequate. This paper proposes a question under discussion based annotation to improve the statistical correspondence. The results show the accuracy of the outputs was significantly increased as regards fidelity, adequacy and grammaticality.