<![CDATA[Text classification can be applied by natural language processing. However, one of the problems found in text classification is sarcasm sentences. Sarcasm can change the meaning of a sentence into the opposite. To solve this problem, the researcher proposes a combination of Fasttext word embedding with the Deep Learning model, namely Bi-LSTM in the case of sarcasm detection in tweets. Fasttext can represent words by utilizing sub-word, so they can obtain information from a word that has never been found and can understand words that have affixes. While Bi-LSTM can study the semantics of words that affect classifying tweets. The experiments were conducted on the use of Fasttext hyperparameters, namely vector size, window, minimal number of word occurrences, epochs, and word2vec model architecture. Based on the experimental results, Fasttext hyperparameters have different effects where there are increasing and decreasing in the value of the evaluation results. Improved evaluation results are more visible by using epochs at a value of 100 and using CBOW architecture.]]>
Copyrights © 2022
