Sentiment Analysis with LSTMs

This repository contains the iPython notebook and training data to accompany the O'Reilly tutorial on sentiment analysis with LSTMs in Tensorflow. See the original tutorial to run this code in a pre-built environment on O'Reilly's servers with cell-by-cell guidance, or run these files on your own machine. There is also another file called Pre-Trained LSTM.ipynb which allows you to input your own text, and see the output of the trained network.

Downloading Data

Before running the notebook, you'll first need to download all data we'll be using. This data is located in the models.tar.gz and training_data.tar.gz tarballs. We will extract these into the same directory as Oriole LSTM.ipynb. As always, the first step is to clone the repository.

git clone https://github.com/adeshpande3/LSTM-Sentiment-Analysis.git

Next, we will navigate to the newly created directory and run the following commands.

tar -xvzf models.tar.gz
tar -xvzf training_data.tar.gz

Requirements and Installation

In order to run the iPython notebook, you'll need the following libraries.

• TensorFlow version 1.1 (Currently not supported for 1.2, 1.3, 1.4 etc but if someone wants to submit a pull request, I'd be open to that :) )
• NumPy
• Jupyter
• matplotlib

Docker

With Docker, you could just mount the repository and exec it.

1. Install Docker. Follow the docker guide.

2. Build docker image

   cd LSTM-Sentiment-Analysis
   docker build -t="@yourname/tensorflow_1.1.0_py3" .

3. Run the container from the image

   docker run -p 8888:8888 --name=tensorflow_yourname_py3 -v /@YourDir/LSTM-Sentiment-Analysis:/LSTM-Sentiment-Analysis -it @yourname/tensorflow_1.1.0_py3

   and visit the URL(http://localhost:8888/)

4. Stop and restart the container

   docker stop tensorflow_yourname_py3
   docker start tensorflow_yourname_py3
   docker attach tensorflow_yourname_py3

   If jupyter is down, relaunch it by using the command below.

   cd LSTM-Sentiment-Analysis
   jupyter notebook --ip=0.0.0.0 --allow-root

Installing Anaconda Python and TensorFlow

The easiest way to install TensorFlow as well as NumPy, Jupyter, and matplotlib is to start with the Anaconda Python distribution.

1. Follow the installation instructions for Anaconda Python. We recommend using Python 3.6.

2. Follow the platform-specific TensorFlow installation instructions. Be sure to follow the "Installing with Anaconda" process, and create a Conda environment named tensorflow.

3. If you aren't still inside your Conda TensorFlow environment, enter it by opening your terminal and typing

   source activate tensorflow

4. If you haven't done so already, download and unzip this entire repository from GitHub, either interactively, or by entering

   git clone https://github.com/adeshpande3/LSTM-Sentiment-Analysis

5. Use cd to navigate into the top directory of the repo on your machine

6. Launch Jupyter by entering

   jupyter notebook

   and, using your browser, navigate to the URL shown in the terminal output (usually http://localhost:8888/)