Development of Compound Non-Standard Word Dataset Using Crowdsourcing Method

During the COVID-19 Pandemic, many activities, such as communication and socialization are forced to be carried out through digital devices. Many communication applications have been developed to improve the users’ communication experience, namely chatbots, auto-translation, sentiment analysis and many other applications. However, most Indonesian use non-standard words when communicating with each other. These non-standard words used caused the text processing software work to be nonoptimal. This research aims to create a corpus of standard and non-standard words. This corpus can then be used to normalize the non-standard word to the standard version. The crowdsourcing method was chosen to create the dataset. This research has successfully collected 371 records as final corpus data. The most common problem is the difference in respondents’ perceptions in determining single or compound non-standard words. There were 5 problems found in the forming of corpus data, namely character repetition in non-standard words, various forms of loan words, typing errors, differences between respondent’s answers, and expression word without standard.