Using text embeddings to display search results spatially.

Building a statistical vector-space model from your corpus of text documents affords many advantages.

Take a search query such as ‘carbonatite’. Using text embeddings (vectors) we can display results not just on a map, but also by how ‘similar’ those locations are to the query. See the associated screenshot.

This allows us to discover locations that may not mention the query keyword, or perhaps even close synonyms. But that location may have similar word patterns leading to the potential discovery of a new less obvious analogue, undiscovered opportunity, locality or train of thought to pursue further.