LRE i-vector participants may find the key for the dev segments under Donwload/Datasets

HELP

Need help ? Please consult the FAQ below, and if you can't find an answer, please contact us.




Using the online platform

 Lockout threshold  

The number of authentication attempts before your account becomes locked is 3. If this happens, an automatic email will be sent to you. Please open it .

 To view your submission status  

After logging in, go to Account->Profile and select view submission profile.

The scoring server will update the status of your submission. Please refresh frequently this page to see the progress of the processing of your submission.

 Attaching a photo to your profile  

NIST does not store any photos on its servers for the i-vector Machine Learning Challenge. Please create gravatar http://en.gravatar.com/ .

Your photo will be automatically associated with your profile on the i-vector website.
 Your submission is not processed  

1) Please make sure that your submission archive contains only one text file using .txt extension. No [sub]directory, no other files.
2) Please make sure that the archive is bz2 archive .

 Join the mailing list  

Please contact us.

 Multiple registrations per organization 

NIST strongly discourages multiple registrations from same site/organization. Each registration should reflect a separate research effort. Please note that participants in the Language Recognition i-Vector Challenge can upload up to 100 submissions per day.



FAQ

SRE i-vector challenge

 Are the i-vectors a mixture of males and females?  

ANSWER: Yes, both development and evaluation sets contain males and females. However, this metadata is not provided.

 Will NIST be scoring cross-gender non-targets?  

ANSWER: Yes, there will be cross-gender trials scored.

 Can the unlabelled development set and the evaluation set be seen as one homogenous database, or can we expect some kind of domain mismatch between development end evaluation sets?  

ANSWER: There is not intentional data set shift included in the evaluation data with respect to the development set.

 Are participants allowed to train system parameters on their own labelled data, or is parameter training strictly limited to the unlabelled data given as part of the challenge?  

ANSWER: Parameter training should be limited to the unlabeled data given as part of the challenge.

LRE i-vector challenge

 Can I use some additional data that I have access to in order to develop or train my system?  

ANSWER: No. We are requesting that participants not use data beyond what was distributed as part of the challenge.

 What is the label I should use in my submission to indicate that a segment is out of set? 

ANSWER: The required label is out_of_set (using underscores in between the words).