Skip to content

Latest commit

 

History

History
35 lines (24 loc) · 1.86 KB

README.md

File metadata and controls

35 lines (24 loc) · 1.86 KB

Generating-TV-Script

This project will generate Simpsons TV scripts using RNNs. Part of the Simpsons dataset of scripts from 27 seasons are used in this project. The Neural Network will generate a new TV script for a scene at Moe's Tavern.

Results

Below are the generated scripts:

moe_szyslak:(sighs) what's the point?... same ol' stinkin' world...(to moe) he seems nice.
lisa_simpson: how'd mindless tester's off my this friend is gonna go tsk, tsk, tsk, tsk, tsk, tsk, tsk till broom.
carl_carlson: i thought you said chug-monkeys. what beverage, brewed since ancient times, is made from hops and grains?
scary him miss_lois_pennycandy:.
lenny_leonard: plus his wife was madonna.
ned_flanders: what're blessing.(small sob)
moe_szyslak: eh, sam: call for the musical in the pledge of allegiance. bugging me.


moe_szyslak: hates me, sam: laugh at the musical in the fridge, the man who spews harmony) pope's beers, you just gotta warn you, they must be the ugliest beer and a wad of bills.


moe_szyslak: hey, hey, hey, hey! freaky.


moe_szyslak:(dumbest glass, sam:) dad?
scary him on, sam: lee is walther hotenhoffer and i'm in the pharmaceutical

The Files

  • check_tensorflow_gpu.py : check tensorflow version and whether your machine have a gpu available.
  • preprocess.py : methods to preprocess, save, load data and parameters.
  • build_network.py : set up the structure for RNN and put it all together by building a tensorflow graph
  • train_network.py : train the model using the structure built in 'build_network.py'
  • generate_script.py : utilize the trained model to generate TV scripts
  • main.py : pass parameters, variables and data to the model, 'main.py' makes the model much easier to train and to tune.