Well Document Extraction Using LLM Prompts for Wellbore Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The extraction of critical wellbore information from unstructured and varied well reports is labor-intensive, time-consuming, and prone to human error, causing delays in the plugging and abandonment process of oil and gas wells.
Innovation Solution
A wellbore extraction tool utilizing generative AI to automate the retrieval of wellbore parameters, converting unstructured data into structured formats, and integrating tailored prompts with Large Language Models (LLMs) for precise data extraction and quality control, presenting the data in interactive tables and well schematics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual extraction methods are used for wellbore information from unstructured reports, then flexibility in handling various report formats is maintained, but extraction time and labor intensity increase significantly
Solution Approach 1:
The patent replaces manual mechanical extraction processes with an automated system combining Optical Character Recognition (OCR) technology and Large Language Models (LLMs). The OCR component converts unstructured report images and PDFs into machine-readable text, while the LLM automatically extracts wellbore parameters through natural language processing, eliminating the need for manual reading and transcription of various report formats.
Solution Approach 2:
The patent introduces an intermediary processing layer consisting of text extraction modules and prompt engineering components. This intermediary system converts diverse unstructured report formats into a standardized text format that the LLM can process, enabling the model to handle various report types without requiring manual adaptation for each format.
2Productivity
If automated extraction tools are implemented, then extraction speed and consistency improve, but accuracy may decrease due to handling unstructured and varied report formats
Solution Approach 1:
The patent implements a feedback mechanism where the LLM generates extraction results that are then validated against the original report content. The system uses prompt engineering to instruct the LLM to verify extracted parameters by referencing specific sections of the source documents, and includes provisions for manual review and correction, creating a closed-loop feedback system that continuously improves accuracy.
Solution Approach 2:
The patent performs preliminary text extraction and structuring before the actual parameter extraction process. The OCR component pre-processes the unstructured reports into organized text formats, and the system prepares standardized prompts with contextual information before submitting to the LLM, reducing the cognitive load and potential errors during the extraction phase.
3Loss of information
If comprehensive data extraction is performed from voluminous reports, then completeness of wellbore information improves, but processing complexity and resource requirements increase
Solution Approach 1:
The patent selectively extracts only the essential wellbore parameters needed for plugging and abandonment operations from the voluminous reports. The system identifies and extracts specific parameters such as wellbore depth, casing dimensions, cementing details, and abandonment specifications, filtering out unnecessary information and presenting only the critical data in an organized format.
Solution Approach 2:
The patent divides the complex extraction process into distinct functional segments: OCR text extraction, text preprocessing, parameter identification prompts, LLM processing, and result validation. Each segment handles a specific aspect of the extraction task, reducing the overall complexity by breaking down the monolithic process into manageable, independent modules that can be developed and maintained separately.
Data Source
AI summary
The present disclosure relates to systems and methods for automatically extracting data from well documents. The systems and methods obtain extracted text from the well documents and create a tailored prompt with instructions for extracting the data from the extracted text. The systems and methods provide the tailored prompt to a large language model. The large language model uses the tailored prompt to automatically extract the data from the extracted text. The systems and methods receive the data from the large language model and present the data in an interactive table.