Daily, a consumer is exposed to hundreds of chemicals used in the formulation of a multitude of products, from cosmetics and personal care products to cleaning products, and even in food.
The ubiquitous nature of these substances and their wide use, determines the observance of strict regulatory requirements at the time of their approval. Thus, companies that manufacture or use chemical substances in their final products are required to carry out the respective toxicological analyses. Currently, these are carried out manually by Safety Assessors, who use different sources to unequivocally identify, characterize and attest the regulatory compliance of each chemical under analysis. The creation of a Toxicological Profile may even require consulting hundreds of pages of information.
Thus, we identified the opportunity to introduce a disruptive solution, capable of improving the efficiency of this process.
The SafetyDesk project addresses this opportunity, taking advantage of the most recent advances in Artificial Intelligence and Natural Language Processing. It is intended to achieve the automation of research and data collection processes in multiple sources of information, as well as the processes of analysis of that information and extraction of relevant data, and considerably speed up the task of building the toxicological profile through the introduction of an assistant intelligent virtual, capable of suggesting texts to be included in that profile, with a high level of technical quality, generated through advanced data-to-text techniques based on the data collected and processed automatically. The construction of the toxicological profile with the support of an intelligent assistant, also intends to guarantee the inclusion of any data relevant to the toxicological analysis, ensuring the highest and unavoidable quality standards.
Objectives and expected results:
• Research and development of a solution that reduces the time needed to carry out toxicological analyzes of chemical substances by at least 70%;
• Implement Artificial Intelligence techniques, namely Natural Language Processing and Machine Learning that allow the solution to make text recommendations to be used by the Security Evaluator.