Automatic identification of irony

Sentiment analysis has been applied to large masses of data produced by social media, allowing to investigate users' opinion about products, brands and news. However, the analysis of text that contains irony remains a challenge, since irony reverses the meaning of a text. This paper aims to detect irony in Twitter posts. For this purpose a dataset was built by crawling ironic and not ironic posts. The construction of the dataset included the creation of features through Bag of words (BOW) and n-grams. The dataset was used to construct a Support-vector machine (SVM) model which was evaluated by K-fold cross-valiation method.