A topic model approach to identify and track emerging risks from beeswax adulteration in the media

The European Food Safety Authority (EFSA) develops methodologies and tools for the detection of emerging risks in food feed. This includes identification drivers risks, such as frauds, which requires innovative approaches. In this study, an unsupervised machine learning technique called Latent Dirichlet Allocation (LDA) topic model, was applied on a media corpus view detecting rapidly specific fraud incidents media, i.e. Europe Media Monitor Medical Information System (EMM/MEDISYS). LDA model can explore large collection documents discovering themes associated with organize summarize text identifying topics comprised them, where is defined pattern words their probability to belong it. As incident, beeswax adulteration taken example. Beeswax be adulterated financial gain, and, although it product from apiculture, might enter chain when introduced honeycomb honey pots. With example, total 2276 news articles were retrieved EMM/MEDISYS classified into 10 showing different levels relatedness adulteration. A manual screening all allowed validate classification made by model. that found most relevant contained indeed reported official sources. addition, those signals potential cosmetic wrapping sectors. remaining highlighted emergence new market opportunities supported identified signals. It concluded used process information support definition more filters direct use stakeholders involved monitoring, assessment management frauds.