FastText and Bi-LSTM for Sentiment Analysis of Tinder Application Reviews

Abstract

Nowadays technology affects all aspects of society, one of the innovations and creativity in the field of technology is the emergence of online dating application media. The application makes it easy for users to find a partner according to their respective criteria. The most popular online dating app is Tinder. The rise of the use of online dating applications invites controversial sentiments in the community. With this problem, a sentiment analysis is needed to find out the opinions and views of users about Tinder. This study proposed the fastText and Bi-LSTM models used to determine the optimization performance of the fastText and Bi-LSTM methods in sentiment analysis and compares the performance of the fastText and Bi-LSTM models with the fastText and Bidirectional Encoder Representations from Transformers (BERT) models. Based on the experiment, fastText and Bi-LSTM produced the highest performance in the 4th fold scenario with 88% accuracy. Based on the comparison of the three model performances, the fastText and BI-LSTM models can outperform the fastText and BERT models on sentiment analysis of user review datasets in the Tinder application.