Thanks to cloud-native technologies, the oil and gas industry is now pursuing how to retrieve information from unstructured data. This is especially relevant in mature exploration and production (E&P) areas such as the Norwegian Continental Shelf (NCS). Valuable information from historical unstructured data (documents, reports, images, etc.) can help when improving processes related to exploring a new area, planning a new well, and creating a rig activity plan. However, the biggest challenge with this unstructured data has always been that it is typically stored in internal servers and within complex folder structures, making the search for a single file a time-consuming -- if not impossible -- process. This paper details a cloud-based methodology of searching and contextualizing unstructured data in order to make it available to domain experts through Application Programming Interfaces (APIs).
We have developed a framework for configurable document processing that makes raw data searchable by leveraging existing cloud-based services such as popular natural language processing (NLP) and computer vision APIs. This makes it possible to extract content and make it searchable and contextualized to related assets (wellbores, seismic, etc.) based on keywords and geospatial location. Machine learning categorization supports file grouping into document types, such as a work order or rig action plan.
To illustrate this methodology we present two examples. The first example shows how a user can draw a polygon on a map to retrieve and extract all the data within that polygon. This can be used in reporting tools or to build knowledge graphs. The data is fully contextualized to the geographical area of investigation. The second example shows content extracted from well reports, providing information about historical issues that happened to offset wells, which supports the planning of a new well in the same area under similar conditions.
Content extracted from unstructured data offers an opportunity to add valuable information to support risky and time-consuming processes such as evaluating an exploration prospect and planning a new well. Searching for and contextualizing this content are the first steps toward making the retrieved information instantly available and more interactive for domain experts.