When it comes to defining depression, doctors usually ask patients specific questions about mood, mental diseases, lifestyles and personal histories, and used these responses to make a diagnosis. Researchers from mit have created a model that can detect depression in people without having to ask them specific questions, starting instead from their natural conversation and writing style.
According to the chief investigator of the project tuki Alhana, “the first hint we get that person happy, excited, sad or is in a very special mood, like depression, because of the conversation. To deploy a scalable model for defining depression, it is necessary to minimize the number of restrictions that you apply to the data. You need to hold a normal conversation and to allow the model to define the human condition in the process of natural communication.”
How to identify depression in a person? The AI will have the conversation
The researchers call their model “uncontextual” because there is no limit to the types of questions that can be asked, or the types of responses that will be heard. Using a method of modeling sequences, the researchers fed the model text and audio from interviews with people in and out of depression. The accumulation of sequences, patterns emerge, such as the natural use of words like “sad” or “down”, and audio signals, which are more monotonous.
“Model sees the sequence of words and identifies these laws as more likely to find people with depression or without it,” says Alkhanai. “Then, if she sees the same sequence in new people, it can identify the presence of depression”. During testing, the model showed a 77 percent success in identifying depression due to what has bypassed all other models, which mostly rely on structured questions and answers.
Perhaps one day the model assistant tool the doctors or even will be the basis for future systems of artificial assistants.
I think that will help? Tell us in our chat in Telegram.
AI from MIT can identify a depression on the conversation
Ilya Hel