Novel language processing system detects mental health distress in employees by analysing their speech

Key potential customers include companies employing content moderators and trust and safety agents

Fionn Delahunty and Mihael Arcan, co-founders of Lua Health

Lua Health is a spin-out from the University of Galway whose founders, Mihael Arcan and Fionn Delahunty, have developed a system that analyses human language to identify early signs of mental health deterioration in employees.

The system works by plugging into an organisation’s existing employee assistance programme (EAP) and when employees interact with the EAP, or communicate with each other via the company’s preferred communications channel, Lua’s AI algorithm will pick up the linguistic biomarkers in their speech that signal mental distress.

“We’re focused on patterns rather than on individual words as when a person’s mental health status is shifting there are discernible changes in their language patterns,” Arcan explains.

“Being able to offer an appropriate intervention in the initial stages of mental ill health makes treatment faster and more effective. Unfortunately, most individuals tend to wait too long before reaching out for help. This is why Lua is focused on the early detection of potential problems and signposting people to the best solution as quickly as possible. We are not providing the solutions ourselves as this area is already well served by wellbeing and mHealth providers. Instead, we identify the beginnings of a problem and link individuals to the treatments already available.”

READ MORE

Arcan is a former research fellow at the University of Galway’s Data Science Institute, where he worked in the natural language processing unit. He specialises in neural machine translation architecture, a subfield of natural language processing that uses computers to translate languages. Delahunty is a psychologist with a background in applied data science, who was closely involved in the development of the Moodpath depression app.

Lua’s system is called the Lua CoPilot. It’s a B2B product charged for on a software-as-a-service (SaaS) basis and aimed at organisations with an EAP already in place. Key potential customers are companies employing content moderators and trust and safety agents as Arcan says they are, “vulnerable to developing mental ill-health due to the nature of their work, which can involve the identification of violent content on social media”.

While it might sound like there is an element of “big brother” about Lua’s CoPilot, Arcan stresses that it is very much an opt-in or opt-out system based on consent and confidentiality. In trials to date employees have seen it as something positive as it shows an active duty of care by their employer in terms of protecting their mental health.

The CoPilot is not linked to any personal employee devices and there is a strict onboarding process that goes through the use protocols in some detail. All personal data is redacted, any cues about individual identity are removed and there is no sharing of information with employers.

Why you need a good night’s sleep — and 12 tips on how to get itOpens in new window ]

The CoPilot can be linked into existing communications platforms such as Teams and Slack and to get things going the system checks in daily with users (through a series of questions) to gauge how they’re feeling in terms of their mental wellbeing. Once the system gets to “know” a user the daily checks stop and the system whirrs away in the background, analysing comments and conversations to pick up any distress cues. If the system recognises a problem it will send a message to the user and they can take it from there in terms of deciding what level and type of help they might need.

Learn another language and keep your brain fitOpens in new window ]

Backing for the Lua Health CoPilot has come from a pre-seed investment round of €500,000 from Growing Capital (an early-stage investment firm focused on supporting Irish-based start-ups) and Enterprise Ireland. Lua Health employs five people. The potential market for the system is international (the CoPilot works at a semantic level so this gets over any language barriers) and the company is now running trials with interested potential customers. Later this year it will launch a seed round of €2.5 million to scale and go global.