desilinguist / ibleu Goto Github PK

Whatever the name of the document I load, the DocId is presented as "fakedoc".

To reproduce:

1. Select hypothesis file.
2. Click "Compare to another system?"
3. Choose a different 1st hypothesis file
4. Click "Compare to another system?" again

Now the option for selecting a second hypothesis file will disappear, and you will have to reload the page in order to load all files.

Hey! I was using ibleu and was wondering how accurate this system is?

I was trying to calculate BLEU score for the Chinese text. I am using a plain text file for the input. I am getting same BLEU score of 84.07 for every lines and the cumulative bleu score of 5.20 or something.

All I want to say is that, it is showing wrong for every text, even if it is right.

I am attaching both the reference & hypothesis files.
GoogleEN-CN.txt
RefEN-CN.txt

It appears that when I compare two documents the scores are switched. Here are some screenshots: