Chat-Bot With Speech Input
Chia-Sheng Chen
Flow
ASR
Text
Magic
Response
TTS
This Topic
"I don't feel good"
"Why do you say you do not feel good?" *
* generated by AI created in 1960s
Model
Attention Is All You Need NIPS '17, A. Vaswan et al.
Baseline
Audio
Transformer
Text
Transformer
Response
Proposal
Audio
Transformer
Response
Model
Model
MFCC
dim = 39
Multi-Head attention
d_model = 39
h = 3
d_k = d_v = 13
Hidden Layer
d_hidden = 64
Experiment
deck
By qitar888
