Klasifikasi Emosi Lirik Lagu dengan Long Short Term Memory dan Word2Vec

Abstract

This research focuses on the classification of emotions in song lyrics using LSTM (Long Short-Term Memory) and Word2Vec embedding. Emotion classification in lyrics plays a crucial role in music recommendation systems, sentiment analysis, and understanding the affective aspects of music. The study explores the effectiveness of LSTM, a type of recurrent neural network (RNN), in  capturing  the  sequential  dependencies  and  patterns  in  lyrics,  combined  with  Word2Vec embedding to represent the semantic meaning of words.The dataset consists of a collection of song lyrics labeled with 2 emotions. The lyrics are preprocessed and convertedinto word vectors using  the  Word2Vec  model.  The  LSTM  model  is  then  trained  on  the  preprocessed  lyrics  data, aiming to predict the corresponding emotion category for a given set of lyrics. Experimental results demonstrate that the proposed approach achieves a maximum accuracy of 72.8% in classifying emotions  in  song  lyrics.  The  LSTM  model  leverages  the  sequential  information  in  the  lyrics  to capture   the   emotional   context   effectively.   The   Word2Vec   embedding   enhances   the representation of words, allowing the  model to  understand the semantic relationships  between words and better discriminate between different emotional categories.

Keywords: TextProcessing, Classification, LSTM, Word2Vec