Utilizing Social Media for Identifying Drug Addiction and Recovery Intervention

This dataset is for the paper "Utilizing Social Media for Identifying Drug Addiction and Recovery Intervention" published in Data Analytics for Smart Health (DASH) workshop (https://sites.google.com/view/ieee-dash-2020) co-located with the IEEE BigData 2020 conference.

Overview of the Paper

Topic-specific social media forums such as Reddit have become popular platforms for users discussing health-related information as well as for scientific analysis of that information. Such discussions among users have been found to be effective in providing useful insights and assistance in many healthcare applications. This study focuses on one such application, where we utilize Reddit posts related to drug addiction/substance abuse, in order to help the addicted persons. We observe some linguistic differences in the posts as users gradually move from the addicted stage to more and more advanced recovery stages. We then classify user-posts on Reddit as to be indicative of drug addiction or of different stages of recovery of the user.  By annotating Reddit posts with the help of standard social and health psychology literature, we develop a Machine Learning classifier based on linguistic features, to classify the posts among different classes related to addiction and recovery. Finally, we identify users having an intention to recover and perform a personalized mentor recommendation by referring them to users who are already in their advanced stage of recovery based on the usage of drugs. To our knowledge, this work is the first attempt to utilize social media for helping addicted users having the intention to recover with personalized mentor recommendations to facilitate their process of recovery.

Overview of the dataset 

The dataset contains posts from drug addiction-related Subreddits which are topic-specific communities within the Reddit OSM (Online Social Media). We considered 100 distinct Subreddits related to drug addiction and recovery, which are listed in the Wiki page of the Subreddit `Drugs' (https://www.reddit.com/r/Drugs/wiki/subreddits). We annotated 3151 posts out of all the posts collected as one of the 5 classes: 'Addicted', ' E(early)-Recovery',  ' M(maintaining)-Recovery',  ' A(advanced)-Recovery', 'Others'. 

The dataset has the following 8 fields, 

* id: A unique identifier for each post
* username: The username for each post
* title: Each post is associated with a title
* body: The main descriptive part of the post is identified by its 'body'
* #comments: This field represents the number of comments the post has received
* score: This field represents the number of upvotes received by each post
* label_classification: The label given to a post for the classification task (one of the 5 classes stated above)
* label_recommendation: The label given to a post for the recommendation task where the 'Addicted' class of classification task was further divided into 'Addicted with intention to recover', and 'Addicted without intention to recover'

Note that the dataset has been anonymized by replacing all personally identifiable information (PII) such as usernames, email-ids, phone numbers, SSNs, etc. by using the tool available at https://github.com/mns-llc/bitsnarf  as well as using some standard dictionary-based methods.