问一下机器翻译中的排序问题
大家好，初次在贵坛发帖，借人气问一下，现在的机器翻译系统中大多关注于对源语言句子的分析，有没有关于对生成目标语言时的句序或语序转换相关方面的文章或研究？谢谢！问一下机器翻译中的排序问题
Generally speaking, in SMT, the reordering of the phrases in target sentences is controlled by the language model and the distortion model. In traditional systems like Pharaoh, the distortion model is just approximated by some exponential function. For the convenience of decoding, usually the distortion is restricted. Some detailed discussion can be found in following papers:Richard Zens, Hermann Ney (2003) A Comparative Study on Reordering Constraints in Statistical Machine Translation. ACL 2003
Shankar Kumar, William Byrne (2005) Local Phrase Reordering Models for Statistical Machine Translation. HLT-EMNLP 2005
And I remember somebody mentioned that the unconstrained reordering is actually harmful, seems it's Phillip Koehn, but I can't find the paper in hand.
In last years HLT-NAACL, there are some papers discussing how to train a distortion model, just like some phrase bigram probabilities. But it's a bit trivial, so I didn't find it's of any interest, it just seems people didn't find any novel direction in SMT :)
But if you are looking for reordering methods in rule-based or example-based MT, it's a different thing, as I'm only familiar with SMT.
ps: I broke my second iMac -_-bb I'm in a shit linux system now. Forgive my not being able to type chinese here.
问一下机器翻译中的排序问题
I really appreciate your reply.：） 呕液！:lol :victory:页:
[1]