Tweets Reporting Abuse Classification Task

In recent decade we have noticed a considerable increase of reports or confession posts of abuse victims on twitter. Most of the times victims does not report it to their guardians or the concerned authorities. Teenagers and minorities are most affected group of abuse. Part of these victims tweet about their incident to let go of pain and suffering or as a cry for help. Identifying such reports are challenging because of unavailability of annotated training data, and high degree of data sparsity. To address this we are hosting TRACT in DSC OMG

• We release the first small scale manually annotated corpus for abuse classification problem
• Propose a shared task for this problem TRACT

Task

This new, multi-class classification task involves distinguishing three classes of tweets that mention abuse reportings: "report" (annotated as 1); "empathy" (annotated as 2); and "general" (annotated as 3)

Important dates

• Registration deadline : 23 June
• Training and Validation data release : 11 June
• Test data release : 26 June
• Predicion submission : 28 June
• Leaderboard release: 30 June
• System description paper submission : 1 July

Data

• Training data : contains 3000 tweets (37 report, 80 empathy, 2883 general)
• Validation data : contains 1500 tweets (19 report, 36 empathy, 1445 general)
• Test data : blind dataset

Evaluation metrics

F1 metrics are considered for evaluation of systems

Submission Guidelines

• Short paper(upto 4 pages) of system and results
• Instructions for authors
• Latex Template Arxiv Preprint

Organizers

1. Saichethan M. Reddy
2. Kanishk Tyagi

Contact

for more information regarding shared task contact organizer

References

• Miriyala Reddy, Saichethan; Tyagi, Kanishk ; Anand Tripathi, Abhay; Deoli, Rajat (2020), “TRACT: Tweets Reporting Abuse Classification Task Corpus”, Mendeley Data, V1, doi: 10.17632/my2vkfyffd.1