Interesting paper published in Computers and Geosciences. “3D lithological mapping of borehole descriptions using word embeddings” Fuentes et al (2020).
Applied to groundwater studies in New South Wales, Australia.
Also an interesting paper published in the AGU. Dictionary‐Based Automated Information Extraction From Geological Documents Using a Deep Learning Algorithm – Qiu – 2020 – Earth and Space Science.
Using text embeddings in some form for geoscience has been around for a while.
I may hold the honour of publishing the very first peer reviewed paper on text embeddings and geoscience (if anyone can find an earlier paper please let me know!). It can be found in the 2015 Knowledge Organisation (KO) Journal, applied to thousands of geoscience company reports.
In 2017 I trained a model using 100 years of papers from the Society of Economic Geology, showing how patterns in that text content may be combined with geohealth data from government databases.
More recently, I made a post last year on the different methods and how they have evolved, which may be useful: