Dataset

Data

Subtask A: The dataset comprises of Urdu Nastalíq tweets elucidating the emotions of people as they describe their activities, opinions, and events with the world. The goal of this dataset is to develop a large benchmark in Urdu for the multi-label emotion classification task.

Subtask B: The dataset comprises of threatening tweets in Urdu language. The detailed processes of dataset collection and annotation are described in this paper. The goal of this dataset is to develop a large benchmark for threatening language detection in Urdu.

TASK A:


Training set: Download

Test set: Download

Test set (with labels): Download

TASK B:

Training set: Download

Test set: Download

Test set (with labels): Download

There are two primary sub-tasks in TASK B.

  1. The first sub-task is to classify the given tweet as "Threatening" and "non-threatening". Table 1 shows the statistics of the first sub-task.

  2. The second sub-task is if the tweet is classified as a "threatening" tweet, then it should be further classified as a threat given to an "individual" or a "group". Check the statistics of the second sub-task in Table 2.


Table 1. Statistics of first sub-task.

Table 2. Statistics of second sub-task.