基于短语统计机器翻译解码算法的研究与实现

Research and implement of phrase-based statistical machine translation decoding algorithm

解码器是统计机器翻译研究的关键部分。在基于短语的统计机器翻译的基础上,结合对数线性模型的思想加入多个特征模型,研究了一种动态规划的柱搜索解码算法。详细介绍此算法在解码器中的具体实现,并对翻译速度和精度作了分析。

Decoder is the important part of statistical machine translation research.Based on phrase-based statistical machine translation,a dynamical programming beam search decoding algorithm is put forward combining multi futures model using log-liner model approach.Introduces the implement of the decoder in detail based on this algorithm,and analyzes to the translation speed and precision at last.