Comparison of Normalization of Indonesian Slang Words Using the FastText & Word2vec Model with the Natural Language Processing Approach

The use of slang words is often used as a means of communication on social media such as Twitter, but it is a problem for certain groups because they are difficult to understand if they are said out of context. This can cause communication to be less effective, especially for those who are not familiar with the slang. Therefore, a word normalization approach is needed to translate words into formal language so that they are better understood by the public. Natural Language Processing (NLP) is a computational technique that analyzes and represents text or spoken language to achieve human-like processing. This research focuses on feature extraction techniques such as FastText and Word2Vec to map words to numerical vectors. The results of testing slang words show that FastText has the highest similarity of 0.9934859978 and the lowest is 0.8928895496, while Word2Vec has the highest similarity of 0.9977979123 and the lowest is 0.0975351095. The time required for FastText for training is 0.432 seconds and for normalization 0.016 seconds, while Word2Vec requires 0.027 seconds for training and 0.006 seconds for normalization.