Skip to content

Deep Learning for Classification of Insincere Text

Notifications You must be signed in to change notification settings

jamqd/SincerelyAI

Repository files navigation

Sincerely, AI

Created by: John Dang, Bill Liu, Joshua Young, and Justin Shao

Introduction

Sincerley, AI is a deep learning project with the goal of identifying insincere text. The model is trained on this dataset from Quora containing over 1.3 million data samples of sincere and insincere text. We used transfer learning in order to train our model based on a research paper released by Google AI Research in March 2018.

Running the Demo

Website Landing Page

  1. To create conda environment run the following command from the repo directory.

    conda create --name quora --file requirements.txt

  2. To activate environment run

    source activate quora

    Note: Users running the demo on a machine with a cuda enabled (Nvidia) GPU should run

    conda install tensorflow-gpu

    for significantly faster inference time.

  3. From the sincerelyAI directory run

    python manage.py runserver

    to run the web app.