Natural Language Processing for anticipating supply chain disruptions to maritime trade
Supply chain organizations have for the most part looked at optimizing decisions such as increasing revenue, minimizing costs and improving process by the use of internal data for quite some time now. However with the advent of Covid-19 in 2020 — the brittleness of the response to extraneous factors has been exposed. Issues such as ports being closed due to Covid-19 outbreaks (eg. Ningbo-Zhoushan port in Southern China) to reluctance of workers to return to manufacturing plants given the speed of spread of the newer Covid-19 variants, there were various factors at play providing supply chain pressures. Currently, we are seeing geopolitical factors such as the conflict in Ukraine also affect the supply chain of commodities such as oil and gas.
When considering the maritime component of logistics it is important to note that shipping lanes pay a major role in facilitating supply chain operations on a daily basis, with around 80% of the volume of international trade in goods being estimated to be carried by sea according to UNCTAD. Understanding the major shipping routes used for global trade is the first step to mitigating problems related to maritime logistics and making adjustments to improve supply chains in many industries.
It is important to understand the constraints about global ship routes before diving into understand how natural language processing can be used to identify disruptions in maritime trade. Firstly, container ships operate similarly to bus routes, moving on fixed routes on paths to predetermined ports. This restricts the possible variations and adjustments a ship can make in its planned route. Secondly, shipping routes are concentrated around ports, with some ports handling more than twice the number of ships than average. One of the most important shipping lanes on the planet is the Strait of Malacca, the shortest route between the Pacific and Indian Oceans. About 40 percent of world trade passes through this strait each year, including much of the crude oil that goes from the Middle East to China.
Apart from the Straits of Malacca, there are some few other major shipping lanes which stand out (caveat -this list is not exhaustive and most have been left out).
For example the 350-mile-long English Channel located between England and France, links the North Sea to the Atlantic Ocean. Approximately 500 vessels travel through the channel each day, making it one of the world’s busiest shipping lanes.
The Panama Canal is another man-made waterway that connects the Pacific and Atlantic Oceans. For ships traveling from the east to west coast of the U.S, this route avoids the more treacherous Cape Horn at the tip of South America or the Bering Strait in the Arctic, and takes off approximately 8,000 nautical miles— or 21 days off their journey.
In 2021, approximately 516.7 million tons of goods were estimated to have passed through the Panama Canal , according to the Panama Canal Authority.
In order for us to measure supply chain disruptions in maritime trade we made use of crowdsourcing data from Twitter with regards to the various shipping lanes throughout the world to identify the sentiment across this lanes on daily basis which allowed us to identify if there were potential disruptions along them. This would allow for proactive engagement from downstream elements in the supply chain, for example ports to identify disruptions from vessel berthing delays etc. and optimize their resources based on the upstream disruptions in the shipping lanes for example. Social media listening in this context provides useful supplier data points in areas such as labor relations, regional governmental restrictions, local news regarding strikes or riots, and weather events that can impact operations of suppliers and may result in supply risk. Further analytics can be built on this data to create early warning indicators. The detailed information from social listening is also made available to allowing for detailed drill-down to analyze and pinpoint exactly where the point of friction lies in the shipping lane disruption based on crowd-sourced data. Advanced techniques such as Named- Entity-Recognition and Topic Modelling are used to understand Variables of Interest and Topics of Interest to provide supply chain managers with the necessary data point to actions and support data driven interventions for disruptions. Given below is a snapshot of a dashboard we developed which showcases the daily aggregated sentiment by some of the major shipping lanes in the world — (in the color code a more positive sentiment is indicated by a number closer to +1 while -1 indicates a negative sentiment and 0 is neutral). The dashboard solution also allows drilling down to understand in detailed as to why the sentiment polarity is flagged in these particular shipping lanes.
