Augmenting training data with syntactic phrasal-segments in low-resource neural machine translation