Sentiment classification by text

This project uses a neural network to classify the sentiment of a review as positive or negative.

See the jupyter notebook

Motivation

Practicing noise reduction and performance optimization of a neural network.

Built With

• Python 3 - Language
• Anaconda - Python Data Science Platform
• Jupyter notebook - Web application that allows to create documents that contain live code

Dataset

List of movie reviews from IMDB

Result

• Final speed: 1.2k reviews/sec
• Final accuracy: 85.9%
• Reviews used: 1k

Getting Started

Prerequisites

1. Download and install Anaconda
2. Update Anaconda

$ conda upgrade conda 
$ conda upgrade --all 

Install

1. Clone and enter into the project's root directory by command line

$ git clone https://github.com/machine-learning-experiments/sentiment-classification-by-text.git

2. Create and activate enviroment

$ conda env create -f enviroment.yaml 
$ conda activate sentiment-classification-by-text 

or

conda create --name sentiment-classification-by-text python=3
source activate sentiment-classification-by-text
conda install numpy matplotlib scikit-learn jupyter notebook bokeh

3. Start jupyter notebook

$ jupyter notebook 

4. Your browser will open showing a list of files, click on the sentiment_classification_neural_network.ipynb notebook file

Author

Lorival Smolski Chapuis

This project was developed during the deep-learning nanodegree from Udacity