As social media platforms like Facebook, Twitter, Instagram, and others have continue to expand, there has been an increase in sadness among users . As a result of this, tech companies and researchers have devised methods for detecting and resolving their users’ mental health difficulties. Now, European researchers have devised a sophisticated algorithm that can detect depression in 9 out of 10 Twitter users. The following are the specifics.
Researchers Develop a Twitter Bot to Detect Depression
An innovative algorithm developed by a team of researchers from London’s Brunel University and the University of Leicester can analyze individuals’ mental states based on their Twitter profiles. To determine whether or not a user is depressed, the system collects and analyzes 38 different data points from their Twitter profile.
The new method was created using two datasets, according to the researchers. One database had the Twitter histories of a large number of Twitter users, while the other contained information about their mental health. The team taught the bot with 80% of the data and trained it with 20% of the data.
When it comes to the algorithm’s operation, the bot first filters out people who have less than 5 tweets, then runs the remaining profiles via natural language processing to look for misspelled terms and abbreviations. The system then uses 38 different data points to evaluate the users’ mental state, including the use of positive and negative phrases, emoticons, and other components.
The researchers were able to achieve an outstanding 88.39 percent accuracy when evaluating the new depression-detecting Twitter bot using the Tsinghua Twitter Depression Dataset. On the John Hopkins University CLPsych 2015 dataset, the bot had a 70.69 percent accuracy rate.
“In machine learning, anything that is above 90% is considered outstanding. So, 88 percent for one of the two databases is outstanding,” said Prof. Abdul Sadka, Director of Brunel University’s Institute of Digital Futures. “It’s not perfect, but I don’t believe any machine learning system can attain 100 percent reliability at this level.” However, the closer you get to the 90% mark, the better,” the Director continued.
According to the researchers, the new algorithm could be incredibly useful in diagnosing mental health disorders among social media users. According to the team, the bot might be expanded to other platforms such as Instagram or WhatsApp in the future, and it could potentially be utilized in criminal investigations.
In a release, Prof. Huiyu Zhou, a Professor of Machine Learning at the University of Leicester, said, “The proposed algorithm is platform-independent, thus it can also be readily expanded to other social media platforms like Facebook or WhatsApp.” “The next stage of this research will be to test its validity in various environments or backgrounds,” he continued. “More importantly, the technology developed as a result of this investigation may be applied to other applications, such as e-commerce, recruitment examination, or candidacy screening.”
