On May 28, Noalwenn Sallée, engineer and geosciences project manager at IFP Energies nouvelles, presented to IFP School students a pilot project that implements IBM's Watson automatic learning solution and introduces our students to the possibilities offered by artificial intelligence to support the work of scientists.
Project objective: To improve the efficiency and relevance of geoscience data collection within a large amount of unstructured scientific literature, using an automatic learning algorithm.
The experiment, applied to the field of oil exploration, consisted of creating a system capable of identifying, within a large number of documents, those which are relevant to finding answers to specific questions, and more precisely those that concern the characterization of the source rock, in a regional geological study.
In practice, scientific publications provide information in the form of text, curves or figures. Two types of automatic learning algorithms were tested: one dedicated to image recognition (Watson Visual Recognition, WVR) and the other to text analysis (Watson Knowledge Studio, WKS). The capacity of the designed workflow is promising, thanks to the good performance of the trained models.