Sentiment analysis of public complaints using lexical resources between Indonesian sentiment lexicon and Sentiwordnet

Public complaints were one of the kinds of public participation and awareness to public service implementation. Information from public complaints can be used by the government to improve public satisfaction. In addition, the government can obtain public sentiment from public complaints either on media social or the official government site. Many kinds of research on sentiment analysis have been done, either used statistical method approach, semantic method approach or both. Statistical method approach was widely used. While semantic method approach being the hot topic recently. On semantic method approach, lexical resource was an important component to classily sentiment on text. Namely Sentiwordnet and Indonesian sentiment lexicon. Currently, Indonesian lexical resources for sentiment analysis has grown. But the lexicon doesn't have polarity score that can be measure emotion on text like Sentiwordnet. Sentiwordnet has been widely used on research in English. In this research, we apply Sentiwordnet to classify sentiment on Indonesian public complaints with accuracy 47% either on media Twitter and 56.85% on the official government website's data. Furthermore, we compare it with Indonesian sentiment lexical and get the accuracy 65.4% on media Twitter and 81.4% on the of Ticial government website