Geomodel exploration with natural language

Geomodels are very complex because of the amount of generated data. The difficulty of their analysis is to pick the meaningful variables and the relevant points of view that help the user to draw conclusions. It is often necessary to sum up, to integrate or to calculate statistics on the geomodelling results to obtain relevant information. Getting the correct display or appropriate transformation are the purposes of this project..

Based on the AskR.ai conversational engine, our web-based prototype lets the user ask a tuned Artificial Intelligence with natural language to get a meaningful output answer:

• a numerical value, directly picked or elaborated from the model values,
• a view or a graph built from the model values.

Our solution is designed as a structured database in which geomodel results are loaded. In a tuning process, the prototype IA is taught in term of geological domain vocabulary and classical business questions as well as usual graphic outputs. The user is finally able to formulate questions and get appropriate answer or result.

The benefit of the developed solution is its accessibility for non-specialists since the questions are formulated in natural language. It can also be used by expert geomodelers to help them interpret their results and produce reporting in a more efficient way. Future work could connect the conversational engine with a standalone web-viewer for the geomodels.