/Sentiment-Analysis-CNN

Project developed in Python 3.5 making use of Keras library (using TensorFlow as backend) to make a model capable of predicting sentiment polarity associated with Spanish tweets.

Primary LanguagePython

Sentiment-Analysis-CNN

Project developed in Python 3.5 making use of Keras library (using TensorFlow as backend) to make a model capable of predicting sentiment polarity associated with Spanish tweets.

Architecture

The architecture of the Convolutional Neuronal Network developed is the one proposed by Kim, Y. (2014)

Alt text

Problem

Model was originally developed to predict Spanish tweets. It was applied to TASS CORPUS using word2vect method developed by Cardellino

Execution instruction

Inside the code, you must replace Train1_x/Test1_x and Train1_y/Test1_y with the corresponding files. In _x files must be appear the words of all of tweets concatenated using word2vect vectors. While in _y files, it must appear polarity associated to each tweet. The version of libraries used are:

  • TensorFlow 1.2.1
  • Keras 2.0.6