speech-to-speech-translation icon indicating copy to clipboard operation
speech-to-speech-translation copied to clipboard

Error in inference (Stage 9)

Open CharviVitthal opened this issue 2 years ago • 1 comments

Hello, I'm running inference scripts in run_baseline.sh (examples/s2s_trans). The 9th stage uses files "en.0.norm.lc.rm en.1.norm.lc.rm en.2.norm.lc.rm en.3.norm.lc.rm" but I don't know how to generate these. There is also a comment in the script: "# put reference text (en.0.norm.lc.rm en.1.norm.lc.rm en.2.norm.lc.rm en.3.norm.lc.rm) under current dir." I couldn't find a way to generate these files or their format in your documentation. Please help me by providing either these files or a way to generate them.

CharviVitthal avatar Feb 13 '23 14:02 CharviVitthal

You can download the following files: fisher_test.en.0(en.0.norm.lc.rm), fisher_test.en.1(en.1.norm.lc.rm), fisher_test.en.2, and fisher_test.en.3, through the following link: [fisher-callhome-corpus/corpus/ldc at master · joshua-decoder/fisher-callhome-corpus](fisher-callhome-corpus/corpus/ldc at master · joshua-decoder/fisher-callhome-corpus). Please note that you must have the necessary copyright permissions when using fisher callhome corpus. By the way, it's important to normalize the text prior to calculating BLEU. This includes converting capital letters to lowercase and removing punctuation.

fengpeng-yue avatar Feb 13 '23 15:02 fengpeng-yue