ALTA logo Language Technology Programming Competition 2019
Home
Task Description
Useful Information
Submission
 
 

2019 Shared Task Description: Sarcasm Target Identification

Basic Task Description

Sarcasm is a form of verbal irony that is intended to express contempt or ridicule. Sarcastic text has been reported to be detrimental to sentiment analysis. The connection between sarcasm detection and sentiment analysis is the target of ridicule since it is this target that receives negative sentiment in the sarcastic sentence.

The goal of this task is, given a sarcastic text, to identify the set of words which are the target of sarcasm. The words are returned as a list with all words in lowercase, where all duplicates have been removed. If such set of words is not found, return a fall-back label "OUTSIDE". For example,
  1. "Your shirt reminds me of my 10-year-old": Target: your shirt
  2. "This is the best film ever!": Target: film
  3. "Oh, and I suppose the apple ate the cheese": Target: OUTSIDE
The assumption in each of these examples is that they are sarcastic. While it is a typical sequence labelling task like many other classic NLP tasks, there has been little work on sarcasm target identification.

Evaluation

We will use Kaggle in Class to evaluate the systems.

Data Files and Submission

We will use Kaggle in Class for this year's competition (look for the ALTA 2019 Challenge). The data files and submission instructions will be provided in the competition website.

In order to access the Kaggle in Class pages, you need to register with this shared task.

Important Dates

Release of training data On registration
Deadline for submission of results over test data 18 Oct 2019
Notification of results 21 Oct 2019
Deadline for submission of system description 4 Nov 2018
Presentation of results at ALTA workshop, Sydney (Australia) 5-6 Dec 2019

 

© ALTA 2019. Competition Organisers.