Advanced Tools Tap Digital Knowledge

October 23, 2019

© botulinum21 / Adobe Stock
© botulinum21 / Adobe Stock

Innovative search techniques optimize digital information access for decision-making

Senior management makes strategic decisions based on reliable information that is assembled by support staff. In the aftermath of the big crew change, the makeup and skill sets of the support staff are vastly different and fall into two categories:

  1. Seasoned subject-matter experts (SMEs) who have read and authored numerous company reports and have many years of hands-on experience.
  2. Millennial and younger engineers who are new hires, have limited field experience and are equipped with standard enterprise search tools.

Of course, the information compiled by both groups will vary due to professional experience and backgrounds. Differences in knowledge bases influence decision-making actions. With many senior SMEs already retired, companies need a better way to retain their company information and transfer expertise to less-seasoned staff.

Preserving company intellectual property through advanced solutions
Innovative systems and software can capture the knowledge of experienced SMEs and provide access to new engineers and support staff. Natural language processing (NLP) is an innovative technology that enables review and retrieval of unstructured digital information from digital textual files. Several industries have successfully used NLP-based virtual advisors to assist staff in querying digital documents. [1] Such methods resemble using SMEs to answer questions instantly with the added benefit of being available to the whole team simultaneously. [2,3]

Enterprise NLP solutions are based on “understanding” the contextual vocabulary and grammar that industry practitioners use. While the understanding appears different inside the algorithms, the practical utility is similar to how a human finds proper solutions for technical problems concerning operating and safety issues. A well-built NLP tool can:

  • Intake the question
  • Extract the intent of the inquiry
  • Thoroughly search phrases and paragraphs in historical digital files
  • Return relevant information to satisfy the intent of the question.

Such capability is worth tens of billions in heavily documented process-driven industries, such as oil and gas (O&G). For many energy, manufacturing and O&G companies, operational efficiency is a high priority; unfortunately, industry practitioners spend 80% of their time looking for answers locked away in company information. [4]

So why are NLP solutions not widely used by the O&G industry? The difficulty lies in designing an NLP tool that works for O&G without being too restricted to a single-use case. Large consulting firms offer information-search tools, but these solutions are designed for broader application across multiple industries. The resulting capacity of the offering is the lowest common denominator. For application in the O&G industry, these “blank slate” tools require the client to provide substantial and expensive time to train the tool. In addition, the total possible value of the finished tool is diminished when the underlying models are not designed for O&G.

Conversely, developing a tool specific to a single O&G application is a tough business model. The market size for a single-use case is comparatively small and, consequently, limits investment opportunities. An effective tool must be focused specifically on the O&G industry but also flexible enough to handle diverse applications within.

© katwijksenieuwe / Adobe StockInnovation meets the needs of the O&G industry
An advanced solution has emerged and is focused on the needs of the O&G business. This tool has had successful commercial application in exploration geology for an oil major. The same tool has had successful application in refinery operations and engineering for a multinational chemical company. The more progressive firms that adopt leading-edge NLP solutions will gain competitive advantages under changing market conditions.

Proving the NLP-based virtual advisor tool. Studies were conducted to qualitatively measure the effectiveness of the NLP-based virtual advisor and compare it to current O&G industry methods. A well-established technical forum, owned by the NLP-based virtual advisor developers, encouraged users to post technical questions regarding upstream and downstream issues. Likewise, the forum compiled answers from industry experts. Users of the forum included operators, engineers and SMEs from around the world (Canada, China, Germany, India, Philippines, Saudi Arabia, South Korea, Taiwan, UK and the USA). In June 2018, the NLP-based virtual advisor was added atop the forum and now has access to the forum’s content to craft the most accurate answers for inquiries. To ensure adequate knowledge and resources that a major international O&G company would have, 6,000 digital unstructured textual files were recorded into the forum before starting the comparative study.

In the proving exercises, SMEs were shown question and answer pairings from the NLP- based virtual advisor. The SMEs concluded that the NLP-based virtual advisor was able to interpret and process the technical nuances of information like an experienced human would. For example, when asked, “What are gasoline product specifications?” the NLP-based virtual advisor found answers relevant to “motor spirit,” a predominantly British vernacular. This association of motor spirit to gasoline was inferred by the system from its own research as would be done by an SME.

Time vs. accuracy
In addition to demonstrating qualitative benefits, this study also investigated the comparative success of the solutions between the forum and the NLP-based virtual advisor. Success was measured in the time required to receive and answer a question, as well as, the accuracy of the retrieved information. The time for a human to answer a forum question averaged over 1.5 day. In many instances, some users browsed through previous posts for answers, because the relevant information was obscured by large quantities of irrelevant information.

By comparison, the NLP-based virtual advisor could retrieve answers within seconds. From interviews with users, it is hypothesized that this drastic reduction in time to receive answers caused an increase in follow-up inquiries on peripheral material and not just rewording of the first question. Getting real-time feedback prompted curiosity to learn more about the subject, thus simulating an informative conversation with a SME.

In addition to reducing the time to receive information, the NLP-based virtual advisor was able to return more accurate answers than supplied by the forum. The median number of responses posted by users to a given question in the forum was two across all samples taken. However, the median number of responses generated by the NLP-based virtual advisor based all forum information was 10 across all samples taken. By returning multiple answers to each question, the NLP-based virtual advisor proved more likely to expose controversy in unresolved issues, complementary information forming a more complete answer and false assumptions held by the user.

© Serge Bertasius / Adobe Stock

Example. When the NLP-based virtual advisor was asked, “What is the flow rate handled by a single buoy mooring (SBM) station?”, it returned an answer relevant to a specific instance of a SBM station. It also returned, “The flow rate handled by SBM varies depending upon the pipeline size,” followed by an explanation of how this occurs. These two answers, when viewed together, indicate that more than one design for a SBM station exists and that either a range must be considered or a particular SBM must be specified. Without the NLP-based virtual advisor, the probability increases that a less-experienced user scouring information will find a single value in a document and end the search with incomplete or misinformation. Both are problematic actions, contribute to using the wrong solutions and create a dangerous situation.

General users of NLP-based virtual advisor demonstrated satisfaction with the generated answers. Whenever a user submitted a question to the virtual advisor tool, they were prompted with the option to post that question to the general forum. Satisfaction with the tool was measured as the likelihood that a user would elect not to post their question to the general forum. It is possible that users had other reasons not to post their questions. Instances of electing to post to the forum generally were nearly 1 in 20 across all samples taken, leading to a satisfaction rate of approximately 95%.

Retaining corporate knowledge
O&G companies have a wealth of digital data and knowledge. Unfortunately, as experienced SMEs retire, the new technical staff will need advanced tools to harvest valued information. NPL-based virtual advisors can be trained to understand the details of the O&G industry. Such methods can time-efficiently review digital documents and provide high-quality information for decision-making. With better tools, less-experienced staff gain more knowledge and are productive.

The NLP-based virtual advisor is available to serve as a virtual personal advisor to technical teams across the O&G industry and give real-time advice based on understanding their own data.

[1] Bogdanov, V., “8 Thought-provoking cases of NLP and text mining use in businesses, February 15, 2019. 

[2] Meyers, Kate, Brown and Meyers, “Insurance companies use NLP technology to analyze text and reduce fraud,” March 19, 2014.

[3] Chickowski, E., TechTarget, “5 Augmented analytics examples in the enterprise,” August 20, 2019.

[4] Larsen, Å. H., Equinor CIO Keynote Address, Society of Petroleum Engineers Digital Transformation Study Group, Digital Transformation Annual Congress, May 10, 2019, Houston, Texas.

Alec Walker is the CEO and cofounder of the artificial intelligence firm DelfinSia in Houston, Texas. He holds an MBA from Stanford Graduate School of Business and a BS degree in chemical engineering from Rice University. Walker has led digital transformation and internal entrepreneurship projects for a variety of leading organizations including Intel, Inditex, AECOM, and GM. He worked at Shell as a technical service engineer in the hydrocarbon refining group, a tech tools software product manager, and a reservoir engineer for unconventional O&G.

Current News

Eni Buys Exmar's Tango FLNG for Use in the Congo

Eni Buys Exmar's Tango FLNG for Use in the Congo

Russia Ups Stakes in Sakhalin Energy Standoff in Sanctions Retaliation

Russia Ups Stakes in Sakhalin Energy Standoff in Sanctions Retaliation

SBM Orders Seventh Fast4Ward Hull

SBM Orders Seventh Fast4Ward Hull

Equinor Makes Move on Rosebank

Equinor Makes Move on Rosebank

Subscribe for OE Digital E‑News

Offshore Engineer Magazine