Comparison of Fasttext and Word2Vec Weighting Techniques for Classification of Multiclass Emotions Using the Conv-LSTM Method

Emotions are one aspect of human psychology which includes the ability to control feelings and behaviors that create ways of self-expression. The classification of emotion in text is an area of linguistic research. The purpose of this study is to solve the problem of how to compare the FastText and Word2Vec weighting techniques with the Conv-LSTM combination to the emotion detection text classification process and how the results of the performance analysis are. In addition, it aims to implement and obtain performance analysis results on comparisons of the FastText and Word2Vec algorithms. The method used is a combination of the Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) methods which are then referred to as Conv-LSTM. The Conv-LSTM combination is usually used for image data, but in this study it was applied to sequential text. Sequential text has its own challenges when viewed from the characteristics of the text in this research dataset which tends to be short. Text classification process in detecting labels on keyword research objects "Anies Baswedan". A total of 98,664 tweet data with 6 multiclass classifications, namely happy, sad, angry, disgusted, shocked, and afraid. The highest accuracy results using the Word2Vec Skipgram method with a Max Pooling of 0.8499. In short, this study shows that the word2vec method is better in terms of accuracy compared to the FastText method.