Skip to main content

Table 1 Model architecture for each setup

From: Dual supervised learning for non-native speech recognition

Setup

M L

M S

M STT

M TTS

1

RNN 3 × 512

RNN 3 × 512

RNN 2 × 1024

Wavenet

2

LSTM 3 × 512

LSTM 3 × 512

LSTM 2 × 1024

Wavenet

3

3-gram

LSTM 3 × 512

LSTM 2 × 1024

Wavenet