The University of Essex and Hivedome Limited KTP 24_25 R2

Commodities traders will soon be able to ask their software questions in plain English, rather than clicking through menus or writing code. The project pairs a university natural language processing team with a commodities trading software company. Currently, traders interact with their data through rigid dashboards and query languages. This limits how quickly they can ask follow-up questions or explore unexpected patterns in pricing, supply, or logistics data. The research aims to build a system that understands conversational requests—such as “show me copper prices for last month compared to the same period last year”—and returns answers instantly. If successful, the work will let traders spend less time wrestling with software and more time making decisions. The underlying technology could also apply to other data-heavy industries where non-specialists need to interrogate complex datasets without a programmer’s help. The project is applied, not fundamental science: it adapts existing natural language processing methods to a specific commercial context, with the immediate goal of making a working product.