Rude-Words Detection for Indonesian Speech Using Support Vector Machine

Hate speech and abusive language detection in Indonesian social media: Progress and challenges

“…As described before, the dataset types that have been used in Indonesian HSAL detection research include text, speech, and images. However, for the speech dataset, [18] , [33] , and [50] do not use audio preprocessing (e.g., compression, filtering, panning and lengthening, noise cancellation, etc.) and directly conducting a features extraction process.…”

“…In [18] , they use the SVM model combined with the Pitch Value feature. This is the same as [18] , [33] also used SVM but with different feature extraction. In [33] , they use pitch, Mel-Frequency Cepstral Coefficients (MFCC), INTERSPEECH 2009, and INTERSPEECH 2010 features.…”

“…This is the same as [18] , [33] also used SVM but with different feature extraction. In [33] , they use pitch, Mel-Frequency Cepstral Coefficients (MFCC), INTERSPEECH 2009, and INTERSPEECH 2010 features. Lastly, [50] use LSTM combined with MFCC, INTERSPEECH 2009, and INTERSPEECH 2010 features.…”

“…While researchers used hate speech from previous works, social media platforms, or hate speech rules/laws, researchers usually used sociolinguistics books for abusive language definition. For example, [33] and [1] used the abusive language definition and characteristics from [56] . 15 We also found that some researchers use another term to express abusive language.…”

“…For research on Indonesian HSAL detection using a speech dataset, the research is initialized by [18] and followed by [33] , [50] . Despite the dataset used by [18] , [33] , and [50] not coming from a social media dataset (they build the speech dataset by manual recording), these studies can be used as a reference for research on social media speech dataset like the speech dataset from YouTube video. Last, for hate speech detection using an image dataset, currently, the research has been done by [3] .…”