search

clab / fast_align

Simple, fast unsupervised word aligner
Apache License 2.0
738 stars 159 forks source link

How to train this model in a new corpus? #50

Open zenRRan opened 4 years ago

zenRRan commented 4 years ago

How to train this model in a new corpus? I use this model in en->ch, but I find this result is not good in fact, so I abtain a large bilingual corpus and want to train this model again.

shiman commented 4 years ago

Why not just re-run the training command, just as you did on the smaller dataset?

duterscmy commented 4 years ago

how did you resolve this problem??

zenRRan commented 4 years ago

并没有解决啊~  

------------------ 原始邮件 ------------------ 发件人: "clab/fast_align" <notifications@github.com>; 发送时间: 2020年9月9日(星期三) 中午1:32 收件人: "clab/fast_align"<fast_align@noreply.github.com>; 抄送: "zenRRan"<824203828@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [clab/fast_align] How to train this model in a new corpus? (#50)

how did you resolve this problem??

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

duterscmy commented 4 years ago

可以python force_align.py -h看下，里面给的命令保存的文件应该就是模型，在新语料上训练就可以，不知道我说的跟你问的意思一不一样

Humanpaladin commented 2 years ago

确实 "result is not good in fact"。 请问问题解决了吗？或是有其他的代替方法吗？