Sarcasm Target Detection

URL: https://github.com/Pranav-Goel/Sarcasm-Target-Detection

Description

This project introduces a manually labeled dataset for the novel task of Sarcasm Target Identification. It focuses on detecting the specific subset of words in a sarcastic text that indicate the entity or situation being ridiculed.

Dataset

The dataset consists of two manually annotated files:

• snippets.xlsx: Contains 224 book snippets labeled for sarcasm targets.
• tweets.xlsx: Contains 506 sarcastic tweets, each annotated to identify sarcasm targets.

Annotation Scheme

• The sarcasm target is a subset of words in the text that refer to the entity or situation being ridiculed.
• If no target is explicitly present within the text, a fallback label of “Outside” is assigned.

Citations

If using the snippets dataset, please cite:

• Joshi, Aditya, Pranav Goel, Pushpak Bhattacharyya, and Mark Carman.
  • Sarcasm target identification: Dataset and an introductory approach.
  • LREC 2018 - Eleventh International Conference on Language Resources and Evaluation.

If using the tweets dataset, please refer to the extended version available at:

• ALTA 2019 Shared Task: http://www.alta.asn.au/events/sharedtask2019/description.html
• Citation:
  • Molla, Diego, & Joshi, Aditya (2019).
  • Overview of the ALTA 2019 shared task: sarcasm target identification.
  • In Proceedings of the 17th Annual Workshop of the Australasian Language Technology Association.

Implementation & Results

⚠ Lack of Information:

• No details on methods, models, or evaluation metrics are provided.
• There is no implementation code for sarcasm target detection beyond the dataset itself.