Not detecting language in large text

[creker]

Example https://play.golang.org/p/qupLXwVQc4m

First example is a large text in English. The library can't produce confident result - confidence is negative. Second example is a couple of sentences from the same text. The result is correct. It doesn't matter in which language the text is. After certain threshold it will always break.

I checked https://github.com/kapsteur/franco that seems to be using the same model and trigrams. It works.

[abadojack]

Thanks @creker. I'm looking into it