The Bank for International Settlements is working with the European Central Bank and Detusche Bundesbank to explore how generative AI could be applied to improve inflation forecasting.
Dubbed Project Spectrum, the programme aims to distil actionable insights from data on consumer prices, which is often raw, high-volume and too unstructured to be processed manually and requires extensive automation to parse the results.
The project will explore how GenAI could be used to autmatically categorise billions of individual product descriptions and price observations to improve inflation nowcasting.
In the initial experiments the algorithms will map data from the European Central Banks's Daily Price Dataset to globally standardised household expenditure categories such as the United Nations' Classification of Individual Consumption According to Purpose (COICOP).
The BIS says the developed algorithms and insights could be applicable beyond Europe: the training dataset covers global retail chains and multiple European languages, and the LLMs and embedding technologies allow for the findings to be extrapolatable even to languages not included in the training set.