LLM Driven Smart Assistant for Data Mapping

Data mapping is a crucial step during application migration and application integration. Data (model) mapping comprises “schema matching” to identify semantically equivalent fields between two schemas and ”transformation logic”, a set of rules for converting data from one schema to another. In current industry practice, data mapping is largely manual, performed by domain experts. We present a data mapping solution powered by Large Language Models (LLMs), offering significant improvements in precision over state-of-the-art (SOTA) methods, along with multiple automation workflows that allow users to provide various input triggers (context) for inferring the mappings. We illustrate this contribution using several representative industrial datasets.