Normalized probabilities: only 1.0 in output values

[aleksandra-miletic]

Hi Adrien,

I am currently testing py3langid and I noticed something strange: the normalized probability values in the output are systematically 1.0. I tested texts of different lengths (1 word to several paragraphs) in different languages. I'm using it with Python. Is this something you noticed before?

Thanks,
Aleksandra

[adbar]

Hi @aleksandra-miletic, thanks for your feedback. You're right, this is a bug.

I changed the formula to normalize probabilities along the way and apparently didn't check it properly.

The change in numpy data type also affects things a bit, and I didn't implement the possibility to write classify(self, datatype='uint32') although I wrote such an example in the readme... I'm going to fix this.

[aleksandra-miletic]

Great, thank you!