Beyond Geocoding - LLM Geo-Embedding

Geocoding is at the heart of finding our place in geo-space. Have you ever looked up an address? Used GPS to navigate somewhere? Utilized Latitude and Longitude to find somewhere? - then you have made use of Geocoding. But Geocoding is a notoriously tedious and expensive computing product to establish and use. So, are we “bounded” by geocoding, or can we do better?

Geocoding uses numbers to determine earth locations and relates them to addresses.
- The numbers make sense to computers and the addresses make sense humans.
So, is there a modern number to semantic paradigm that is better suited for modelling the translation of number-related locations to large human language semantics and embedding relative meaning per the “closeness” of locations?? Yes!! - Large Language Modeling (LLM) with semantic search.

LLM maps semantics (sentences with meaning) to a grid of numbers, and positions semantics in related proximity. Does that sound familiar - semantics, grid, proximity? Yes - addresses are semantic sentences, lat-long form grids, and proximity is associated both in address numbers and relative locations of addresses. The process of associating semantics with the grid is called “embedding”.

Could an LLM Geo-embedding prove equal or superior in - data processing, sense-making, and computation cost?