翻译软件是怎么根据拉丁字母判断何种语言的？

https://github.com/dsc/guess-language

Based on guesslanguage.cpp by Jacob R Rideout for KDE which itself is based on Language::Guess by Maciej Ceglowski. Original repo is at Google Code; repackaged with package metadata here.


东亚语言之类，可以用 unicode 分区判断字符，拉丁类的语言，直接搜常见组合，比如德语的白名单： https://github.com/dsc/guess-language/blob/master/guess_language/trigrams/de

统计学吧

理论上现在翻译软件多多少少都会用到算法
那就可以识别语种了吧

可以通过 N-grams 去判断，大多数 NLP 工具都是这样。