An example conversation in the English conversation corpus
Synthesized samples
The following samples are unseen during training.
Sample 1
Context
B: as an author!
B: i’m quite talented.
B: my literature professor at university used to encourage me.
A: this is not a school, mrs lee!
A: i got in touch with you to speak to you about our marketing office!
Current utterance
A: that’s what i’d like to hire you for!
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “that’s what i’d like to hire you for.” | “that’s what i’d like to hire you for.” | “that’s what i’d like to hire you for.” | “that’s what i’d like to hire you for.” |
Sample 2
Context
E: linda is a vegetarian, aren’t you, linda?
B: yes, yes, i am.
E: so is it possible to have some wheat free snacks and some that are suitable for vegans as well?
I: well, i’m not sure.
I: i’m just a technician.
Current utterance
I: i repair the machines.
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “i repair the machines.” | “i repair the machines.” | “i repair the machines.” | “i repair the machines.” |
Sample 3
Context
B: and the office is very comfortable.
A: do you have any promotion prospects?
B: no, i don’t think so.
A: do you attend training courses?
B: yes, sometimes.
Current utterance
B: what do you do?
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “what do you do?” | “what do you do?” | “what do you do?” | “what do you do?” |
Sample 4
Context
A: hi michelle, sandra!
C: how are you?
C: wonderful!
C: all together again, just like the old days!
C: ok guys, what shall we do?
Current utterance
C: i want to have lots of fun!
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “i want to have lots of fun!” | “i want to have lots of fun!” | “i want to have lots of fun!” | “i want to have lots of fun!” |
Sample 5
Context
B: merchandise was destined for china!
E: china?
E: excuse me, um are you the owner of the blue moon in orange massachusetts?
B: no i’m not!
B: i work for spectre, we’re a multinational sporting goods manufacturer.
Current utterance
B: i’m speaking with fast shippers, aren’t I?
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “i’m speaking with fast shippers, aren’t I” | “i’m speaking with fast shippers, aren’t I” | “i’m speaking with fast shippers, aren’t I” | “i’m speaking with fast shippers, aren’t I” |
Sample 6
Context
A: the others will be here in a minute.
A: let’s get started straight away.
A: right.
A: before you know it we’ll have cleaned and tidied everything!
B: you haven’t changed anne, cleaning, dusting, and tidying!
Current utterance
B: it’s just like when we were at university.
Synthesized speech
| Vanilla FastSpeech 2 | With GRU-based context modeling |
With DialogueGCN-based context modeling |
(Proposed) With MSRGCN-based context modeling |
|---|---|---|---|
| “it’s just like when we were at university.” | “it’s just like when we were at university.” | “it’s just like when we were at university.” | “it’s just like when we were at university.” |