Skip to content

My tensorflow implementation of "A neural conversational model", a Deep learning based chatbot

License

Notifications You must be signed in to change notification settings

LEchenhao/TwitterQA

 
 

Repository files navigation

Deep Twitter Q&A

Introduction

Have you ever wanted to know how someone special would reply on Twitter? Train a model on their twitter data.

This repository has adapted Conchylicultor/DeepQA to specifically ONLY work with tweets. Note that by default scraping Twitter and using their API will only yield you a maximum of 3200 tweets; most likely less. Twitter data will be collected by using the Twitter API. More specifically, all the answers of the person of interest in respond to questions is what the model will be trained on.

If there are companies out there who would like to have a custom QA application, feel free to contact me.

Usage

Fill in twitter credentials in chatbot/credentials.json.

Then you can run:

python main.py --twitter_name gvanrossum

to build a model after the BDFL.

... 20 minutes later ...

Run

python main.py --twitter_name gvanrossum --test interactive

to have an interactive QA session with Guido.

Example

Question asked after having trained on default settings:

Q: which editor do you use ?
A: emacs of course !

Twitter data

You can see the collected twitter data at:

data/tweets/<username>-answers.txt
data/tweets/<username>-questions.txt

Installation (quoting Conchylicultor)

The program requires the following dependencies (easy to install using pip):

  • python 3.5
  • tensorflow (tested with v0.9.0 and v0.11.0)
  • numpy
  • CUDA (for using gpu, see TensorFlow installation page for more details)
  • nltk (natural language toolkit for tokenized the sentences)
  • tqdm (for the nice progression bars)

Further instructions

You're advised to experiment with the possible paramters to make it a better model.

Have a look at the original repo for more information.

About

My tensorflow implementation of "A neural conversational model", a Deep learning based chatbot

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 51.3%
  • CSS 36.4%
  • JavaScript 10.1%
  • HTML 2.0%
  • Shell 0.2%