Skip to main content

Table 5 Comparison of the different QbE STD evaluations

From: Search on speech from spoken queries: the Multi-domain International ALBAYZIN 2018 Query-by-Example Spoken Term Detection Evaluation

Evaluation

Language/s

Type of speech

# queries dev./test

Primary metrics

MediaEval 2011

English, Hindi, Gujarati, and Telugu

Tel.

64/36

ATWV

MediaEval 2012

2011 + isiNdebele,Siswati, Tshivenda, and Xitsonga

Tel.

164/136

ATWV

MediaEval 2013

ALB, BAS, CZE, NN-ENG, ISIX, ISIZ, ROM, SEP, and SET

Tel. and mic.

> 600/ > 600

ATWV

MediaEval 2014

ALB, BAS, CZE,NN-ENG, ROM, and SLO

Tel. and mic.

560/555

C nxe

NTCIR-11 2014

Japanese

mic. workshop

63/203

F-measure

NTCIR-12 2016

Japanese

mic. workshop

120/1620

F-measure ATWV MAP

ALBAYZIN 2012

Spanish

mic. workshop

60/60

ATWV

ALBAYZIN 2014

Spanish

mic. workshop

94/99

ATWV

ALBAYZIN 2016

Spanish

mic. workshop+parliament

102/106+95

ATWV

ALBAYZIN 2018

Spanish

mic. workshop+BNews+conv.

102 + 103/106 + 108 + 91

ATWV

  1. Tel. telephone, mic. microphone, BNews broadcast news, conv. conversational, dev. development, ATWV actual term-weighted value, Cnxe normalized cross entropy cost, MAP mean average precision, ALB Albanian, BAS Basque, CZE Czech, NN-ENG non-native English, ISIX Isixhosa, ISIZ Isizulu, ROM Romanian, SEP Sepedi, SET Setswana, SLO Slovak