Analysis of Indonesian Sentiments Using Indonesian Sentiment Lexicon by Considering Denial

Abstract

The amount of public opinion can be very much against a trendy thing. Seeing this, it is not possible to read every opinion manually to get a conclusion from all opinions because it will take quite a lot of time. Sentiment analysis can be used as a solution to the problem. However, public opinion is not always written in just one sentence, it can even be written in more than one sentence. It should also be noted that sentences in Indonesian are divided into two based on their forms, namely simple sentences and compound sentences. Simple sentences only consist of one clause, while compound sentences consist of one or more clauses. Each clause in compound sentences can have differences based on whether there is denial the clause or not. Seeing this, this paper seeks to improve performance by considering the denial, sentence form in Indonesian, and punctuation that is used as a paused or finished intonation. Data used in this paper obtained from Twitter during the campaign of the election in West Java 2018. The methodology of model building consist of 4 stages: data preprocessing, feature extraction, classification, and evaluation. The model resulted in this paper has improved accuracy by 3% from 60.15% to 63.7%.