A multidimensional approach for detecting irony in Twitter

摘要: Irony is a pervasive aspect of many online texts, one made all the more difficult by absence face-to-face contact and vocal intonation. As our media increasingly become social, problem irony detection will even pressing. We describe here set textual features for recognizing at linguistic level, especially in short texts created via social such as Twitter postings or "tweets". Our experiments concern four freely available data sets that were retrieved from using content words (e.g. "Toyota") user-generated tags "#irony"). construct new model assessed along two dimensions: representativeness relevance. Initial results are largely positive, provide valuable insights into figurative issues facing tasks sentiment analysis, assessment reputations, decision making.