Opensource GEOAssist v3.0 Released! Harness the power of natural language with LLM-driven map creation, utilizing 1.6 million mineral and rock localities from Mindat, the world’s largest minerals database.
New functionality added this weekend. The example natural language query shown in GEOAssist is “Show copper, gold, antimony indium, niobium, beryllium, tantalum and tungsten in Peru”. The LLM used by GEOAssist interprets the query and interacts with the API of the mindat database to show these occurrences spatially. There is an option to download the spatial mineral data shown in the map as CSV, to integrate with other data using applications such as Spotfire, QGIS or ESRI ArcGIS for visualisation and spatial analytics.
It treats ‘copper’ as a commodity, which means any locality containing a Cu-bearing mineral will be included e.g., malachite, azurite, chalcopyrite, bornite not just native copper. As well as native elements, you can query on individual minerals (e.g. sphalerite, stibnite) and rocks (e.g. pegmatite).
This intuitive way of interacting with associative information can potentially open up databases and lead to numerous possibilities for making new connections and insights. This builds on existing functionality in GEOAssist for autonomous Internet literature search & PDF download and reference gathering, automated Geoscience Knowledge-Graph (GeoKG) creation from PDFs, plate tectonic reconstruction and lithostratigraphic unit summaries.
The excellent mindat database is a project from the Hudson Institute of Mineralogy. Over the coming weekends by popular demand, I’ll add an OCR option for the PDFs so those PDFs which are just scanned images which are downloaded can also be exploited. Link to the GitHub software code of GEOAssist V3.0 here: https://github.com/PCleverleyGeol/GeoAssist—An-open-source-autonomous-research-agent-for-geoscience-data-and-literature.-/
A Geodyssey - Geoscience Text Analytics and Enterprise Search Research 
Leave a comment