A blog post co-authored by Amazon’s Carlos Castro and Kevin Wang explains how well drilling reports can be analyzed with natural language processing (NLP). Amazon Comprehend, its own brand NLP service, was used in an automated pipeline to extract insights from unstructured data in well activity reports.
US operators report well activity to the Bureau of Safety and Environmental Enforcement (BSEE) using Form 133 which includes a free text field where unusual events can be recorded. NLP was used to search for key words in the text area that might indicate, for instance, a well control problem. Other key entities can then be captured in an engineering database.
The apparently straightforward task involved a pipeline built with AWS Step functions to coordinate multiple services into ‘serverless workflows’, create a training dataset and train the system. A ‘state machine’ leverages AWS Lambda functions containing application logic. A ‘Split SNWAR’ function, Elasticsearch, Amazon DynamoDB and Kibana also ran (and we have skipped quite a bit of the pipeline). If you are tempted to try Amazon’s labyrinthine NLP offering, read Castro and Wang’s blog. On the other hand, you could try grep.
© Oil IT Journal - all rights reserved.