Skip to main content

Table 6 False alarm rates for OOV words when using Kaldi-based ASR with N-gram LM

From: Dynamic out-of-vocabulary word registration to language model for speech recognition

 

Position

Freq.

2 mora

3 mora

4 mora

5 mora

ONE

0

0

0

0

ALL

.011

.008

0

0

500r

Random

500

.005

.012

.002

0

1kr

Random

1000

.019

.011

.004

.001

2kr

Random

2000

.019

.008

.004

.002

5kr

Random

5000

.032

.011

.003

.004

500c

POS

500

.009

.008

.002

.001

1kc

POS

1000

.009

.008

.002

.001

2kc

POS

2000

.015

.008

.002

.001

5kc

POS

5000

.038

.013

.002

.002