Frequently Asked Questions (FAQs)
Q1: Is there a deadline for registering the teams?
A1: No deadline for registration.
Q2: In Subtask C for predicting the hate-speech type, It says
"<Hate Speech Type, ex: HS1> #One of HS types: HS1 (Race), HS2 (Religion), HS3 (Ideology), HS4 (Disability), HS5 (Social Class), and HS6 (Gender)."
Is NOT_HS a possible tag in this subtask? Or does the provided test set for this subtask only include instances of hate speech? This will affect how the model is trained.
A2: For subtask C, you will have a tweet from all test tweets, and you are asked to predict its hate speech type which can be NOT_HS
of course. If the type is HS1 and you predict it as HS2 for example, this is considered as an error in subtask C. Of course, this is a correct answer for subtask B.
Q3: I could not find any tweets for 'HS4' in the training set (Subtask C). Is this right?
A3: Yes, HS4 appeared 0 times in the training data, and it appeared 1 time in the dev set.
Insulting people based on their disability or diseases was very rare in our collected data set. Maybe in other data sets, it has a different distribution.
Q4: I'm having trouble with encoding the datasets, the text is shown in strange.
A4: You can open the following links in new tabs in your Internet browser (ex: Google Chrome), then save them to your local drives (right-click/"Save As..."):
training data link: https://alt.qcri.org/resources/OSACT2022/OSACT2022-sharedTask-train.txt
development data link: https://alt.qcri.org/resources/OSACT2022/OSACT2022-sharedTask-dev.txt
In addition, in the CodaLab competitions, two buttons are added to download training data and dev data as zipped files. You will find them under "Participate/Files": Public Data, and Starting Kit.
Q5: Should I participate in all the three Subtasks? I want to participate in only one for example. Is this allowed?
A5: You can participate in only one Subtask or two of them. It's not required to participate in the three Subtasks.