forked from Snowdar/asv-subtools
-
Notifications
You must be signed in to change notification settings - Fork 0
/
results.txt
31 lines (22 loc) · 2.2 KB
/
results.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Obtained by Snowdar, Zheng Li at XMUSPEECH in May 2020.
AP20-OLR challenge sets three tasks that will be evaluated and ranked separately.
Task 1: Cross-channel LID is a close-set identification task, which means the language of each utterance is among the known traditional 6 target languages, but
utterances were recorded with different channels.
Task 2: Dialect identification is a open-set identification task, in which three nontarget languages are added to the test set with the three target dialects.
Task 3: Noisy LID, where noisy test data with the 5 target languages will be provided
Baseline results on AP20-OLR-ref-dev (to help estimate the system performance when participants repeat the baseline systems)
------------------------------------------------------------------------------------------------------------------
Task[Cavg/EER%] [Kaldi]i-vector [Kaldi]x-vector [Pytorch]x-vector
------------------------------------------------------------------------------------------------------------------
Cross-channel LID 0.2965/19.40 0.3583/36.37 0.2696/26.94
Dialect identification 0.0142/1.33 0.0267/2.53 0.0302/3.13
------------------------------------------------------------------------------------------------------------------
Baseline results on AP20-OLR-test (standard test set for the challenge)
------------------------------------------------------------------------------------------------------------------
Task[Cavg/EER%] [Kaldi]i-vector [Kaldi]x-vector [Pytorch]x-vector
------------------------------------------------------------------------------------------------------------------
Cross-channel LID 0.1542/19.40 0.2098/22.49 0.1321/14.58
Dialect identification 0.2439/23.94 0.2370/22.25 0.1938/19.74
Noisy LID 0.0967/9.77 0.1079/11.12 0.0715/7.14
------------------------------------------------------------------------------------------------------------------
Please refer to https://speech.xmu.edu.cn/ or http://olr.cslt.org for more info about the OLR Challenge 2020 and on how to request the challenge data used in this recipe.