Social Media data for exploring the association between Cannabis use and Depression

The tweets from the time period of January 2017 to February 2019 are collected from Twitter using Drug Abuse Ontology (DAO) terms related to Cannabis and depression lexicon. A sample of 9888 tweets is manually annotated by domain experts and research students under the supervision of domain experts. The domain experts who are co-authors of this data are substance use epidemiologists with experience in Interventions, Treatment, and Addictions Research.

The annotation scheme has the following coding:

1. Treat: Cannabis is used to help/treat/cure depression.

2. Cause: Cannabis causes depression or makes symptoms worse.

3. Others: Implies other types of relationships, or too ambiguous/unclear to interpret.

4. Addiction: Lack of access to cannabis leads to depression, showing potential symptoms of addiction.

5. Ambiguous: If word related to weed/cannabis or other cannabis products like dabs or edibles is missing or if the mention of depression-related words is missing.

We will continue to add and manage the dataset with new annotations and versions using machine learning techniques like weak supervision and transfer learning. For core computer scientists, this dataset can be useful while working on NLP problems like Relation extraction, Entity disambiguation, Relation prediction, etc. For interdisciplinary experts, this social media dataset can be explored to understand the association between cannabis and depression.

We acknowledge the contribution of Jason Roden and Melissa Wahl from Wright State University for participating in the coding under the supervision of domain experts. This project is sponsored by the National Institute on Drug Abuse (NIDA) Grant No. 5R01DA039454-02 to the Ohio Center of Excellence in Knowledge-enabled Computing (Kno.e.sis) and the Center for Interventions, Treatment and Addictions Research (CITAR) titled: Trending: Social media analysis to monitor cannabis and synthetic cannabinoid use. Any opinions, findings, conclusions or recommendations expressed in this material are those of the investigator(s) and do not necessarily reflect the views of the National Institutes of Health.