We share research datasets with researchers for non-commercial academic use. Each dataset requires a signed data use agreement. Select a dataset below, review the terms, and submit a request. Approved requests receive a download link by email.
Instagram Cyberbullying Dataset
Instagram media sessions — each an image together with its comment thread — labeled by human annotators for cyberaggression and cyberbullying. The labeled set contains 2,218 media sessions drawn from public Instagram profiles. Each session has at least 15 comments, and the sessions were selected to raise the rate of bullying instances. Annotators recruited through CrowdFlower saw the image and its comments together and labeled each session using a definition of cyberbullying that requires repeated aggression and an imbalance of power.
The data supports research on detecting and understanding cyberbullying in media-based social networks, including the roles of image content, comment timing, language and social-graph features.
Data use agreement
By requesting access you agree to the following terms:
- Use the data solely for non-commercial research purposes.
- Do not attempt to identify, contact or re-identify any individual in the data, or otherwise invade the privacy of the users represented in it.
- Do not share or redistribute the data; others who want it must submit their own request.
- Store the data securely and delete it when your research is complete or if the lab asks you to.
- Cite the associated publication(s) in any work that uses the data.
Request access
Sign agreement & request access
Requests are reviewed by the lab. If approved, the download link is sent to the email address you provide.
How to cite
Homa Hosseinmardi, Sabrina Arredondo Mattson, Rahat Ibn Rafiq, Richard Han, Qin Lv, and Shivakant Mishra. Analyzing Labeled Cyberbullying Incidents on the Instagram Social Network. Social Informatics (SocInfo 2015), LNCS 9471, pp. 49–66, 2015.