This repository used RNNs to generate Seinfeld TV scripts ("Fake") based on the data from the scripts for past 9 seasons.
-
Get the Data
-
Explore the Data
-
Implement Preprocessing Functions:
- Lookup Table
- Tokenize Punctuation
- Pre-process all the data and save it
-
Build the Neural Network
- Check Access to GPU
- Input Batching
- Test the Dataloader
- Model
- Defining Forward and Backpropagation
-
Neural Network Training
- Train Loop
- Hyperparameters
- Training
-
Generate TV Script
- Generate Text
- Generate a New Script
Generated Scripts:
jerry: what?
jerry: yeah, yeah. i know, but, uh, i think it's really a very unusual idea. i can't go to the bathroom.
elaine:(sarcastic) oh... i know, you know the guy, i just wanted to know how you can get a little nervous.
george: oh, yeah, but i was wondering if i can be able to be a little bit of a big salad, but i got to get some popcorn.
jerry: i don't know, i know, i can't believe it was my business, but he was in the middle of a new york position and i have no idea.
george: you know, you know, it's not really a good thing. i just got a message to do that.
jerry: i don't know, i don't know what i'm gonna do.
george: i can't do that.
george: i think it's a little unusual.
george: you know...
george: well, what do they do.
jerry: well, i guess i could do this. i mean, i got it.
elaine:(to george) what?
kramer:(to jerry) hey!
jerry: oh, hi jerry. i know, i just don't want a little bit.
jerry:(to jerry) what are you talking about?
jerry: yeah.
george:(to jerry) you know, it's a good idea.
jerry: oh, yeah, yeah.
jerry:(to kramer) what about you?
george: i know.
george: i think it's like a woman.
george: well, i was just wondering if i'm gonna have a chance to go.(he leaves)
elaine: hi.
jerry:(on intercom) yeah, i got a little problem with you.
george: oh, no, no, i can't.
jerry: well, i guess i could be able