Applying Generative AI (ChatGPT) to subsurface and wells content. I’ve been applying OpenAI’s ChatGPT to the graph data structures produced by the OpportunityFinder(R) and GeoClassifier(R) NLP algorithms after they have processed unstructured data (reports, presentations, literature etc) at a sentence level.
This technique called prompt engineering, allows you to apply Large Language Model structure but on your own content rather than the Internet.
Pre-processing the text can focus on the most important information, not necessarily just the ‘top 10’ documents from a search index, which misses outliers.
The example attached summarises the petroleum system element risks for a basin, based on input across many documents. However, a whole range of questions can be supported.
These can be represented visually, see Gen AI radar charts below on two different areas. Get an ‘opinion’ automatically from text on where risks lie!
We can also surface contradictions in reports to stimulate new ideas, questions and areas for further investigation. See the example below, ChatGPT on OpportunityFinder(R) output.
A Geodyssey - Enterprise Search & Discovery and Text Mining Research 
Leave a Reply