Evolving from physical to semantic models
A well-defined physical model with constraints provides a solid foundation for subsequent transformations or analytics using SQL or external dashboarding tools. But Snowflake gives users much more flexibility than traditional BI and allows users to “talk to their data” using natural language. Using Cortex Analyst, Snowflake’s fully managed LLM-powered assistant, users bypass SQL altogether and get direct answers to their analytics questions.
However, let’s not forget that LLMs, at heart, are non-deterministic probabilistic pattern predictors—and relatively expensive ones at that. The more hints or context we can provide to the model, the less guessing it will have to do, thereby maximizing its accuracy. To achieve this, we transform the physical model into a format that the LLM can interpret: a semantic model.
As described below, a semantic model has wide-ranging benefits beyond Cortex and...
