Constructing A Corpus Of Figurative Language For a Tweet Classification and Retrieval Task

摘要: Twitter is an intriguing source of topical content for tasks involving the detection phenomena such as sarcasm and metaphor. The hashtags that users employ to self-annotate their own micro-texts can often facilitate targeted retrieval texts with desired characteristics. Though tweets tagged #sarcasm are highly likely be sarcastic, lack a topic model sarcastic makes it difficult detect when tags used in expected way, or indeed, retrieve not explicitly this way. In study, we explore how tweet-retrieval classification system benefit from constructing task-specific corpus, irony, metaphor detection.