sentiment-spanish

LAST VERSION CHANGES:

From version 0.0.25 we no longer use tensorflow. The model is now based on Naive Bayes classifier.

How does it work?

sentiment-spanish is a python library that uses Naive Bayes classification to predict the sentiment of spanish sentences. The model was trained using over 800000 reviews of users of the pages eltenedor, decathlon, tripadvisor, filmaffinity and ebay. This reviews were extracted using web scraping with the project opinion-reviews-scraper

Using the rate in the user reviews we trained the model to learn from the language in them. We achieved a validation accuracy (accuracy over fresh data, no used for training) of 90%. For more details regarding the training of the neural network model check the repo sentiment-analysis-model-neural-network

Why?

I believe there are not many solutions to sentiment analysis in spanish based on neural networks.

Install and use

First to install the package

pip install sentiment-analysis-spanish

Import the package

from sentiment_analysis_spanish import sentiment_analysis

run the sentiment analysis:

sentiment = sentiment_analysis.SentimentAnalysisSpanish()
print(sentiment.sentiment("me gusta la tombola es genial"))

you will see that it outputs

0.9304396176531412

For instance if you use the text

sentiment = sentiment_analysis.SentimentAnalysisSpanish()
print(sentiment.sentiment("me parece terrible esto que me estás diciendo"))

you will see that it outputs

2.1830853580533075e-06

which as you see is very close to 0.

Output and meaning

The function sentiment(text) returns a number between 0 and 1. This is the probability of string variable text of being "positive". Low probabilities mean that the text is negative (numbers close to 0), high probabilities (numbers close to 1) mean that the text is positive. The space in between corespond to neutral texts.