LSTM-based for Spanish and English Sentiment Analysis

Abstract

Machine Learning (ML) and Deep Learning (DL) are techniques widely used to program artificial intelligence (AI) because they are based on given knowledge from a specific set of information. This study focuses mainly on applying natural language processing (NLP) by predicting the sentiment of the sentences given in the Spanish language. One of the most significant contributions in sentiment analysis that this study has made is the provision of a new dataset for further research and comparing the study with an English dataset with tweets with positive and negative labeling. The Spanish dataset consists of 18,000 comments filtered and preprocessed, and the English dataset consists of 20,000 tweets already preprocessed for this study LSTM model’s input. English today is now the most widely used language around the globe, and a considerable amount of research has already taken place. Thus, making this dataset and analytical comparison a very abundant resource for a computer scientist to have a point of comparison for the performance of the LSTM model. This study compares the results using four machine learning algorithms: (1) support vector machine, (2) random forest, (3) Gaussian Naïve Bayes, (4) gradient boosting, and a deep learning algorithm such as LSTM model in English and Spanish. This study also shows the five models mentioned above being employed in the testing dataset created. Of these 5 models, the LSTM model has the highest reliability, with 95 percent precision in Spanish and 99 percent precision in English. The first four machine learning algorithms achieved 99 percent accuracy on the Spanish dataset and even better on the English dataset.