Multi-Rule Based Ensemble Feature Selection Model for Sarcasm Type Detection in Twitter

Table 1

Sample n-gram and skip gram feature extraction.

Unigram	{‘it’}	{‘was’}	{‘supposed’}	{‘to’}	{‘be’}	{‘a’}	{‘joke’}
Bigram	{‘it’, ‘was’}	{‘was’, ‘supposed’}	{‘supposed’, ‘to’}	{‘to’, ‘be’}	{‘be’, ‘a’}	{‘a’, ‘joke’}
Trigram	{‘it’, ‘was’, ‘supposed’}	{‘was’, ‘supposed’, ‘to’}	{‘supposed’, ‘to’, ‘be’}	{‘to’, ‘be’, ‘a’}	{‘be’, ‘a’, ‘joke’}
1-skip 3-grams	{‘it’, ‘was’, ‘supposed’}	{‘it’, ‘was’, ‘to’}	{‘was’, ‘supposed’, ‘to’}	{‘was’, ‘supposed’, ‘be’}	{‘supposed’, ‘to’, ‘be’}	{‘supposed’, ‘to’, ‘a’}	{‘to’, ‘be’, ‘a’}	{‘to’, ‘be’, ‘joke’}