Sentiment Analysis on different types of texts

In this project, I have implemented a sentiment analyzer that is able to classify different types of input text (tweets, movie reviews, sentences).

Installation

pip install git+https://github.com/leyresv/Sentiment_Analysis.git
pip install -r requirements.txt

Usage

To try the classifier on your own data, open a terminal prompt on the root directory and introduce the following command:

python main/main.py

Models

Here you can find the different models and the code to train them.

Tweets classifier

The tweets classifier has been trained using the Twitter Samples dataset from NLTK. I have extracted the positive and negative frequencies of the words on each tweet and used them as features to train three different classification models using a Logistic Regression, a Naïve Bayes and a Support Vector Machine algorithms. I only keep the best performing one, which is the Logistic Regression classifier, with an accuracy of 0.997 on the test set.

Visualization

Here you can find some notebooks explaining the training process for each classifier.