Unleash the Potential of Upstream Data Using Search, AI and Computer Vision
Hasan Asfoor, Dalal Abadi Alharbi · 2022
Abstract Data is at the heart of Digital Transformation whether it is generated by legacy systems, a word processor, or state-of-the-art sensors. This makes it harder for Upstream professionals to find the right information for operational needs and making decisions. In this work, we describe an approach that utilizes Enterprise Search, AI and Computer Vision to construct a single efficient search layer, called USEARCH, over multiple data repositories. This enables Upstream professionals to perform searches using simple business language and get the information they need wherever it resides. Our approach to build USEARCH consists of five steps. First, establishing a search infrastructure. Second, indexing content and metadata of documents to make them easily searchable from a single layer regardless of the source repository. Third, utilizing Artificial Intelligence to intelligently tag information within data and documents such as wells, reservoirs and business process labels. Fourth, applying Computer Vision techniques to extract tabular information from documents. Fifth, developing an intuitive user interface to simplify finding data. It can intelligently set the business context based on user domain such as Exploration, Drilling or Petroleum Engineering. We applied our approach on over 3 million documents of different types such as drilling reports, reservoir studies, well test analysis, biostratigraphy and geological maps and stored on multiple repositories. This results in a massive improvement. Upstream professionals can now find data through a single layer and thus eliminate the overhead of switching searches between repositories. In addition, multiple search queries that used to take minutes have now been replaces by a single query that takes a few seconds and even milliseconds in some cases. Furthermore, search results are displayed in a business context with direct links to data at the source repositories. The application of Computer Vision allows any tabular data within documents to be exported to databases for fast analysis. Our approach connects all different data repositories and provides user with an intuitive user interface to find data. This unleashes the potential of Upstream data and empowers Upstream analysts to spend more time using data than finding it.