aditya-joglekar / fs02_scoring_toolkit Goto Github PK

Hi, 

We are preparing the submission for ASR1 and ASR2 tracks and I have a doubt regarding the track names.

From the website https://fearless-steps.github.io/ChallengePhase2/ASR.html, I understand Track 1 is non-diarized audios (expected format - single file), Track 2 - with reference segments (expected json files with timestamps)

However, from https://github.com/AJoGaTech/FS02_Scoring_Toolkit/tree/master/egs/sys_results/ASRIt looks the opposite: Track1 has jsons, Track 2 - a single file.

I believe github has more correct information (as we have Track 2 Transcriptions in a single file). But just in case, could you please confirm my understanding is correct? Thanks! 

Hello,

Just wanted to confirm.
Currently, we have some words with hyphen ("-") in reference annotations.
Such tokens are currently fully removed by the normalization.

I see 3 options, but not sure which one is the best:

• replace hyphen with space
• remove hyphens
• keep words with hyphen

Not sure which one is the best, but I believe some of these would be useful to unify the input from the participants