Project Introduction
Personalized responses are essential for having an informative and human-like conversation.
Because it is difficult to collect a large amount of dialogues involved with specific speakers, it is desirable that a dialogue generation model can learn to generate personalized responses simply from monologues of individuals.
In this work, we proposed a novel personalized dialogue generation method which reduces the training data requirement to dialogues without speaker information and monologues of every target speaker.
Model Structure
Generated Sentence Examples
| Input Message |
I saw him talking to her. |
| (seq2seq) |
Have you even seen her? |
| Monica |
I told her he was killed… |
| Joey |
I told you to stop. |
| Chandler |
I know what the truth is. |
| Phoebe |
I just saw you, you know… |
| Ross |
I told you. |
| Rachel |
I’m sorry. |
| Input Message |
You have got to listen. |
| (seq2seq) |
No! It’s my plan! |
| Monica |
Listening. |
| Joey |
What’s wrong? |
| Chandler |
You listen to me! |
| Phoebe |
You’ll understand. |
| Ross |
Listen to me. |
| Rachel |
What are you doing here? |
| Input Message |
We don’t need the price tag. |
| (seq2seq) |
The most of the hat is not true. |
| Monica |
I don’t think so. |
| Joey |
I don’t know. |
| Chandler |
I know that. |
| Phoebe |
I don’t need that. |
| Ross |
I don’t care. |
| Rachel |
I don’t care. |
| Input Message |
Do you like him? |
| (seq2seq) |
No, just stop. |
| Howard |
No, I’m sorry. |
| Raj |
He’s a good kid. |
| Leonard |
No. |
| Sheldon |
He’s a really good friend. |
| Penny |
He’s a cool kid. |
| Amy |
Yeah yeah I know. |
| Bernadette |
He’s my man. |
| Input Message |
Want some coffee now? |
| (seq2seq) |
Let’s go. |
| Howard |
Let’s go. |
| Raj |
Yeah. |
| Leonard |
Come here, please. |
| Sheldon |
No. |
| Penny |
No, no, no, no… |
| Amy |
Come on, I’m going to sleep. |
| Bernadette |
Come on! |
| Input Message |
I think we should tell her. |
| (seq2seq) |
Okay. Well, I’ll see the picture. |
| Howard |
I think I’m going to tell her that. |
| Raj |
I don’t want to… |
| Leonard |
No!! |
| Sheldon |
I know! |
| Penny |
But I don’t know her. |
| Amy |
I don’t know… |
| Bernadette |
I swear it’ll be over. |
- Mixture of multiple datasets
(Trump’s monologues, Starwar movie scripts, Friends, Clinton’s monologues, singer/song-writer’s lyrics)
| Input Message |
Forget it, man. |
Where are we now? |
He’s so cute. |
| Trump |
You’re the one who’s got my daughter! |
Out on this near town. |
He’s…he’s not a great guy. |
| Starwar |
I’m not sure. |
We are on the east side. |
He is not a man… |
| Friends |
I’m not sure one would hurt. |
We are on the street to unk, and we both know that we will play. |
He’s a bad guy. |
| Clinton |
I’m sorry. I’m not sure what you mean. |
We are here. |
He’s a good man, and he’s my friend. |
| Singer |
I don’t know, I was just thinking about him. |
We want to go into town. |
He is so cute! |
Notes
This website is associated with the paper “Personalized Dialogue Response Generation Learned From Monologues” by Feng-Guang Su* and Aliyah R. Hsu* (* indicates equal contribution). If you’re interested in knowing more, please refer to the paper for more details.