Abstract
This paper presents a novel approach to compute food composition data forIndian recipes using a knowledge graph for Indian food (FKG.in) and LLMs. Theprimary focus is to provide a broad overview of an automated food compositionanalysis workflow and describe its core functionalities: nutrition dataaggregation, food composition analysis, and LLM-augmented informationresolution. This workflow aims to complement FKG.in and iteratively supplementfood composition data from verified knowledge bases. Additionally, this paperhighlights the challenges of representing Indian food and accessing foodcomposition data digitally. It also reviews three key sources of foodcomposition data: the Indian Food Composition Tables, the Indian NutrientDatabank, and the Nutritionix API. Furthermore, it briefly outlines how userscan interact with the workflow to obtain diet-based health recommendations anddetailed food composition information for numerous recipes. We then explore thecomplex challenges of analyzing Indian recipe information across dimensionssuch as structure, multilingualism, and uncertainty as well as present ourongoing work on LLM-based solutions to address these issues. The methodsproposed in this workshop paper for AI-driven knowledge curation andinformation resolution are application-agnostic, generalizable, and replicablefor any domain.