Correcting Classified Activities with Natural Language Processing

Abstract For Drilling Contractors and Oil & Gas operators, daily drilling activity reporting is crucial to understand the underlying efficiency of the drilling process, identify risks and uncertainties at the well level and field level, look for opportunities to improve average rate of penetration and to be within the target drilling budget and target time. All the previous while increasing the safety of the people and assets and have the lowest impact in the environment. Daily drilling reports help drilling contractors and operators to understand what, where and when an activity was performed, helps managers to understand teams and contractors performance, and helps executives have a holistic view of the drilling process, and how it meets predefined Key Performance Indicator targets. Analysis does not only look into visible non-productive time, but extends into invisible non productive time. Examples of visible non productive time include drilling borehole assembly failure, critical rig equipment failure, etc. Examples of non visible non productive time include premature pulling out of hole of drilling equipment assuming high bit wear, variations between crew experience during drilling resulting in lower rate of penetration, etc. The challenge the industry faces is that daily drilling activity reporting; along with real time reporting, provides the foundation for drilling analytics. And inaccuracies in reporting the drilling activity, such as reporting the wrong activity, would lead to inaccurate analytics and insights. To overcome this challenge, operators standardize drilling activities reporting by defining drilling codes and subcodes that are used by Drillers, Tool Pushers and Company men to map a drilling activity to a usable code and subcode, and in turn, those codes and subcodes can be used to extract structured data out of the drilling reports for reporting purposes.