A prototype tile server that adds a similarity field to Macrostrat tiles based on legend text similarity
with an search term.
Matching is done with SBERT embedding models. It is possible to use either an out-of-the-box model or a custom model trained on mining reports.
This work was done as part of the the DARPA CriticalMAAS program. This proof-of-concept could be applied to CDR tile layers as well.
- Initial modeling by Meng Ye at SRI.
- xDD embedding model was trained by Ian Ross.