ABSTRACT
The covid-19 pandemic has been pushing the development of online learning systems in Indonesia. In online learning, computer-based essay tests and assessments have an essential role. Essay test systems are designed to mimic the concept of essay tests without being computer-based. The answer from the lecturer is compared to the response from the student. The TF-IDF (Term Frequency -Inverse Document Frequency) cosine similarity is used. It is one of the methods of information re-gathering systems. The process in this model consists of two types: 1) creating a corpus/ inverted file, and the second is cosine similarity (CS) for calculating the similarity of the user's answers with the lecturer's. Creating a corpus/inverted file involves several stages like data collection, parsing sentences into terms, stoplist, weighting with IDF, and term weighting using TF-IDF. The cosine similarity process consists of parsing users' answers, weighting users' answers using TF-IDF, and finding cosine similarity values of users' answers with lecturers' answers using the vector space model. The highest cosine similarity value is taken to give the user's answer points. Testing the Essay Test system produces excellent grades. The tests were done Mean Squared Error (MSE) values resulted in an average MSE value of 3.28 from three students. © 2022 IEEE.
ABSTRACT
Twitter is one of the social media used in Indonesia to express opinions/opinions. One of them is the opinion about Covid-19 which is taking the world by storm. The government's provisions regarding Covid-19 itself reap many pros and cons on social media, one of which is Twitter. In this study, 'Covid-19' will be used as a keyword to conduct sentiment analysis. Sentiment analysis is the process of understanding, extracting and processing textual data automatically to obtain information contained in an opinion sentence. The Naive Bayes Classifier method is used to classify and calculate the total accuracy of the class that has been obtained. Based on the results from the kaggle dataset, there are a total of 2269 tweet documents with the keyword 'Covid-19' on March 23-May 14, 2020 which can be trusted because the data has been labeled by experts. The Naive Bayes Classifier method has 2269 data sets, then divides it into 1815 training data, and 453 data testing data and produces an accuracy of 0.674. © 2022 IEEE.