AI Based Unstructured Document Geotagging Solution

In the oil and gas industry, extracting relevant information for new projects often involves scrolling through a large unorganized or semi-organized unstructured electronic repository of unstructured documents inside an enterprise. However, the digital availability of these unstructured documents and their proper characterization with respect to E&P domain objects remain a significant challenge. To address this issue, we developed an - "AI-based unstructured document geotagging solution" that utilizes text mining, optical character recognition (OCR), natural language processing (NLP), and named entity recognition (NER) to extract entity relations for spatial E&P objects and document category (for example well completion, log data print, daily drilling report etc.) from documents content autonomously. We employed popular language models spaCy, Roberta-base, GPT2 and GPT3.5 and compared the results. The extracted information was then linked to a structured database that contained geo-locations. The solution was developed using the company specific data sciences platform but is capable of being deployed for an operator on-premises or in the cloud. The end-to-end solution can be integrated with any GIS-based dashboard solution, allowing user to select objects from the map and the model can predict related documents to the objects along with document category. This solution will allow user to obtain an organized list of related documents out of electronic repository with unorganized collection of uncharacterized documents, which will allow the user to prioritize and deploy only content of interest to be ingested to the insight and analytics system often hosted in cloud environment.