Detection of Sarcasm Sentences in Indonesian Tweets using SentiStrength
Rajnaparamitha Kusumastuti, Ema Utami, Ainul Yaqin · 2022
Detection of sarcasm sentences in the Natural Language Processing (NLP) field has a fairly high level of difficulty because it ignores elements of facial expressions and intonation of speech style. SentiStrength works by way of unsupervised learning so it is easier to use to classify sentiments more quickly. This study uses SentiStrengthID which has been developed for Indonesian texts. The tweets used in this study are tweets in Indonesian. SentiStrength will classify sentences into 3 classes, namely positive, negative, and neutral sentiments to detect the spread of sarcasm sentences within each sentiment group. The results of this study indicate the SentiStrength accuracy value of 54.52% in detecting sarcasm sentences in Indonesian-language tweets. SentiStrength managed to divide the detected sarcasm sentences into as many as 13 sentences on positive sentiment, 24 sentences on negative sentiment, and 70 sentences on neutral sentiment.