Overview of the 2019 ALTA Shared Task: Sarcasm Target Identification

Overview of the 2019 ALTA Shared Task: Sarcasm Target Identification Diego Moll´a Department of Computing Macquarie University diego.molla-aliod@mq.edu.au Aditya Joshi CSIRO Data61 aditya.joshi@csiro.au Abstract We present an overview of the 2019 ALTA shared task. This is the 10th of the series of shared tasks organised by ALTA since 2010. The task was to detect the target of sarcastic comments posted on social media. We intro- duce the task, describe the data and present the results of baselines and participants. This year’s shared task was particularly challenging and no participating systems improved the re- sults of our baseline. 1 Introduction Sarcasm is a form of verbal irony that is intended to express contempt or ridicule. Sarcastic text has been understood to be a challenge to sentiment analysis because a sarcastic text may appear to be positive on the surface but is intended to be neg- ative. Empirical results also show that sarcastic text is detrimental to sentiment analysis (Maynard and Greenwood, 2014). Applications where senti- ment understanding is important are also impacted by sarcastic text. These applications include di- alogue systems where the correct prediction of sentiment is important to generate appropriate re- sponses. Towards this, computational sarcasm has gained interest in the research community. Sentiment in a text can be understood to be composed of valence (positive/negative) and the target (Liu, 2012). The connection between sar- casm detection and sentiment analysis is the tar- get. Sarcastic text bears a target of ridicule. It is this target that receives negative sentiment in the sarcastic sentence. The goal of the 2019 ALTA Shared Task is the automatic detection of sarcasm targets. Section 2 describes the general aims of the ALTA shared tasks, and the specific aim of the 2019 shared task. Section 3 briefly presents related work. Section 4 describes the data. Section 5 shows the evaluation results. Section 6 presents the results, and Sec- tion 7 concludes this paper. 2 The 2019 ALTA Shared Task The 2019 ALTA Shared Task is the 10th of the shared tasks organised by the Australasian Lan- guage Technology Association (ALTA). Like the previous shared tasks, it targets university students with programming experience, but it is also open to graduates and professionals. The general objec- tive of these shared tasks is to introduce interested people to the sort of problems that are the subject of active research in a field of natural language processing. Depending on the availability of data, the tasks have ranged from classic but challenging tasks to tasks linked to very hot topics of research. There are no limitations on the size of the teams or the means that they may use to solve the prob- lem. We provide training data but participants are free to use additional data and resources. The only constraint in the approach is that the processing must be fully automatic — there should be no hu- man intervention. As in past ALTA shared tasks, there are two cat- egories: a student category and an open category. • All …

Content truncated to 500 words