AI automates geoscience data management

2019 AI Paris conference hears from Total and Amayas Consulting on deep learning-based approach to document classification.

Florian Bergamasco (Total) and Dayron Cohen (Amayas Consulting) presented on the application of artificial intelligence in automating the management of geoscience data. Total’s geoscience data is growing at around 20%/year, the company now has around 15 petabytes of seismic data. Last year Total bought Maersk Oil with a considerable data integration challenge. Following brainstorming sessions and workshops between Total’s data managers and data scientists, a Data Lab was established. The Lab first developed its own expert system internally before going to third parties (LumenAI and Amayas) to enhance and industrialize the solution.

The result is data package classifier using OCR to extract metadata from reports and performing data validation, for duplicates, checking that navigation data is there etc. Amayas’ metadata extractor applies deep Learning, OCR, and image recognition, notably for company logo detection. An 88% accuracy is claimed in the extraction of well metadata from reports. The system works on PDF documents, text or scanned images, all routed through different processes. A list of ‘bigrams’ (pairs of words) is used to score and identify a page. The deep learning runs on a Linux blade on the network. Xception and ResNet also ran. More from AI Paris.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.