Employing BERT model backed by expert knowledge to extract from textual media event of interest along container shipping supply chain

Abstract

Container is the keystone of multimodal supply chains. As container shipping involves numerous actors and because of immense volumes, associated data is teeming. IoT now enables us to see through this mist at the container level. We therefore, we propose a demonstration service to extend visibility by utilizing insights offered by IoT data inherent to containers. The location of containers serves as a starting point to gather information about higher-level circumstances. We armed the service with machine learning algorithms for detecting events of interest along the supply chain through textual exogenous data. An automated information extraction methodology based on BERT model backed by expert knowledge has been implemented. It is illustrated here on a use case to detect climatic events along tracked container route by retrieving tweets from twitter API.