Today, social media is much-used to express emotions, opinions, views, and personal preferences. What, if there was a way to detect if there is any underlying condition behind social media posts of individuals? Now, there is one. A new machine learning algorithm devised by a team of computing scientists at the University of Alberta can identify early signal of depression via written texts such as social media posts.
“The result of the study is to be able to build useful predictive models that identify depressive language accurately,” said a graduate student at the university who devised the model to detect linguistic clues in everyday communication.
While, the model is used to detect depressive language on social media channel Twitter, it can be easily used for text from other domains for identifying depression.
Model uses Writing Samples of Individuals identified as Depressed
The model for English language is developed using samples of writing of individuals, who were identified as depressed on online depression forums. Thereafter, the machine learning algorithm trained to detect depressive language in tweets.
“Meanwhile, this is the first study that shows depressive language has a specific way to be represented linguistically,” said the associate. The model demonstrates that it is possible to detect it, transfer it and further use for depressive language identification tasks.
There are many potential applications of the model, noted the associate. This is from detecting early signs of depression to help clinicians monitor the effectiveness of treatment for their patients over a period of time.
“The algorithm could be integrated with a chatbot that converses with seniors and can label signs of depression and loneliness,” the associated added. The monitor of messages of high-school students to detect if they are suffering from depression could be another potential application of the algorithm.