Communications of the ACM
Bot Can Spot Depressed Twitter Users in 9 of 10 Cases
The bot considers 38 distinct factors, including a user’s use of positive and negative words, the number of friends and followers they have, and their use of emojis, to make a determination on that user’s mental and emotional state.
Credit: Brunel University London (U.K.)
An algorithm developed by researchers at Brunel University London and the University of Leicester in the U.K. can ascertain a person's mental state by extracting and analyzing 38 data points from their public Twitter profile.
The researchers trained the bot on two databases containing thousands of users' Twitter histories, and additional data about their mental health.
It excluded all users with fewer than five tweets, then corrected for misspellings and abbreviations in the remaining profiles using natural language software.
The algorithm identified depression with 88.39% accuracy in one of the datasets, and 70.69% in the other.
Said Brunel's Abdul Sadka, “It's not 100% accurate, but I don't think at this level any machine learning solution can achieve 100% reliability. However, the closer you get to the 90% figure, the better.”
From Brunel University London (U.K.)
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found