Visual classification technology provides energy companies with strategic competitive advantage by making well log archives electronically searchable

DrillingWell_280John Martin, CEO of BeyondRecognition, LLC (“BR”), today announced, “Energy companies are now using BR’s visual similarity technology to convert their paper well log archives into searchable electronic document collections so they can identify wells that may have become commercially viable since having been capped in.”

According to Martin, “Our energy clients tell us that our ability to quickly and accurately convert paper well-log archives into meaningful electronic document collections gives them a strategic competitive advantage in identifying potential new hydrocarbon reserves using their existing well data.”

Martin continued, “For decades, oil companies sank thousands of wells looking for oil or gas that could be produced using then-current technology. If they ran into shale or commercially unrecoverable hydrocarbons, they capped the wells and moved on. Well logs and other well-related documents were placed in storage boxes with minimal indexing and moved to warehouses or salt mines where they have remained untouched. The same thing happened when energy companies acquired other energy companies – the historical well information was simply placed in archives without deriving any competitive insight from that information. Some energy companies have literally millions of boxes of well-related documents from wells they drilled or acquired, but they have had no way to select the ones with useful information.”

BR’s technology now enables energy companies to identify which documents from those archives to migrate to an electronic content management system like Documentum, FileNet, or OpenText, and to decide which data elements to extract from each type of document, e.g., API well numbers, well names, operators, etc.

BR’s visual similarity technology clusters visually similar documents, and only a small fraction of the documents have to actually be examined by data processing personnel. For example, deciding how to process one daily drilling report or one drilling parameters report extends to all such reports, greatly leveraging the time of the document review teams. Individual coding decisions can apply to tens of thousands of documents. Document conversion or migration projects that would require years can be completed in months and with greater accuracy and completeness than was feasible with manual coding.

Martin explained, “‘Convergence’ is the point at which nearly all documents being processed fall into previously established clusters, meaning that they require no further manual review. The larger the collection the more significant convergence becomes. In one recent energy project we processed a test batch of approximately 18,000 documents and used the resulting cluster information for subsequent processing. The first ten segments of approximately 200,000 pages per cluster had a convergence rate averaging 90%. There were under 19,000 new clusters identified in those first 2 million pages, and coding or reviewing those new clusters involved looking at only the first few documents in each cluster. After about 2.4 million pages the convergence rate hit 100%.”

Because BR using visual similarity technology to cluster like documents, even essentially non-textual documents like well plats can be grouped together. Visual similarity also enables BR to identify documents that are visually the same (e.g., the same document that had been scanned twice) as well as those that are bit-for-bit the same (e.g., two copies of the same Word file). Identifying duplicative files can remove 60% of some collections, resulting in significant savings in ECM storage costs.

Comments are closed.