Experts from the Information Processing Center – National Research Institute (OPI – PIB) have developed a system that detects symptoms of depression in social media. The tool turned out to be the best in the international competition Detecting Signs of Depression from Social Media Text.

The award in the competition is not the first success of specialists from the Intelligent IT Systems Laboratory. They have been dealing with natural language processing for many years. They are the creators of Annobot – a platform supporting the creation of systems for detecting emotions in a text. It is an innovative tool for labelling and creating datasets by talking to a chatbot.

“Our solution was based on neural language models, namely RoBERT models. In addition to using the standard model for the English language, we also trained our own model on the specially prepared dataset. It consisted of about 400 thousand emails – mostly about depression, anxiety and thoughts of suicide. The combination of these two models gave the best result”, says the main creator of the system, Rafał Poświata.

OPI PIB experts have long recognized the potential of neural language models. The institute has even created a Linguistic Engineering Laboratory where IT professionals build intelligent knowledge discovery tools from large corpora of text and web data. It was they who developed and implemented the Unified Anti-plagiarism System, which is used by all promoters of diploma theses in Poland.

According to the World Health Organization, approximately 322 million people worldwide suffer from depression. Prompt diagnosis and treatment can significantly improve patients’ quality of life and, in some cases, even save their lives. Therefore, IT experts create solutions that enable quick detection of symptoms of depression, e.g. on the basis of texts published in social media.

Arkadiusz Słomczyński