Safer workplace with AI
Occupational safety is extremely important to any company on the planet. Every workplace accident costs to a company in many ways. At worst, the costs for a single, serious injury can be hundreds of thousands of euros. Let alone the suffering of the injured employee. Improving the occupational safety is an essential part of the strategy in many industries and companies are ready to invest in it. This is when Artificial Intelligence steps in.
Tireless AI finds risks from the written text
SSAB and Outokumpu, the international customers of Sofor, want to be the forerunners in the occupational safety. They have been part of our "Intelligent Safety" project that Business Finland funded in their AI Business program. In the project, we studied how AI can be applied to the collected occupational safety data i.e. to the accident, hazard and near-miss situation descriptions. The aim of our project was to analyze the written descriptions using natural language processing (NLP) techniques. The ultimate goal was to find risk factors which could be attended to and thus prevent future accidents. Instead of a human being, a tireless machine is more suitable for analyzing the huge amount of text.
During the project, we concentrated especially on analyzing the Finnish language, which covered 90 percent of texts. Rest of the texts, including over 20 languages, were combined to the data with machine translation. As a result, we found ways to cluster thousands of descriptions into a manageable set of accident groups without any human (i.e. subjective) hypotheses or restrictions. After the clustering, one could find different employee profiles dominating different accident types, e.g. young workers were prone to different accident types than older employees. As a consequence, one could personalize the occupational safety actions to fit everyone better to gain more effectiveness.
Faster than a human – a machine alerts on time
Injury prevention can be improved, if the near miss reports were automatically analyzed and thus the lag between the hazard and the action could be reduced. For example, one could already at home receive a mobile alert about a slippery parking lot at work. Further, the property manager might not have even read the hazard description, when the workers have already got the instructions from the AI to move the excess packaging causing a danger of stumbling next to the pressing machine. As a result, the anticipation may have prevented the breakage of the knee ligaments and a few months sick leave. In this way, using intelligent methods, one can make the workplace safer. Once again, a company is a step closer to an accident-free workplace.
Head of Analytics