Digital Energy Journal- When natural language beats visualisation

Join | Existing Member Login

You are Home » News » View Article

When natural language beats visualisation

Thursday, October 9, 2014

Computers usually communicate data to us using complex visualisations - while we usually communicate with our colleagues using language - by speaking to them. Could computers communicate with us using language too? By Dr Robert Dale, CTO and Chief Strategy Scientist, Arria NLG

But when it comes time to make use of the insights that might be gleaned from the data, and from charts and graphs that display that data, it's almost always the case that the key information will be communicated from one person to another via language.

Human language has evolved over many tens of thousands of years as a finely-honed tool for conveying information. It is, quite simply, the best means we have for communicating meaning.

It's only recently that machines have started to acquire this capability, in the form of what we might call 'articulate analytics', the automatic generation of narrative text, mixed with graphs and tables, that explains key information and provides a recommendation based on software-embodied expertise.

Our company Arria NLG has been deploying this technology in the Gulf of Mexico, where our Natural Language Generation software is used to provide exception-based surveillance.

An alarm is triggered when the data from a sensor breaches an important threshold, and this spurs the Arria NLG Engine into action.

The Engine analyses all the related sensor data to detect significant events, and it incorporates historical information about previous alerts and the relevant service history.

Based on this analysis, it writes a detailed situational analysis report and provides a recommendation for action.

The software can produce such a report within 60-90 seconds of the alarm. That's in contrast to the 3-4 hours it might take a human expert to produce the same analysis and recommendation.

The technology doesn't replace human engineers, but it makes them vastly more productive.

The engineer can choose to accept or refine the system's report and recommendation, in a tiny fraction of the time that would be required to create it manually from scratch.

Technology
Natural Language Generation, the underlying technology that makes this possible, is a sophisticated combination of tried-and tested techniques from data analytics, artificial intelligence and computational linguistics.

It can take a diverse set of data sources, analyse that data using processes that emulate human reasoning, and then express the results of that analysis and reasoning in natural language, including English, Spanish and Chinese.

So it takes data and turns that data into actionable plain language reports that provide timely decision support because they are easily understood.

It's a multistep process that first considers the range of data available to determine what important information it contains; then, it reasons about how best to present that information in clear and concise language to convey what's significant and why, and what should be done about it.

Human plus computer
To make this work, we capture in software the ability of the engineer to articulate both the problem and the answer that is needed to deal with the problem.

By capturing the knowledge and experience of the engineer in the software, we get easily replicable and scalable expertise that consistently produces articulate output.

Combining automated data collection with the automated knowledge and experience of your subject matter experts means that you are far less likely to miss the vital data required for making the right decisions in the field.

NLG doesn't 'dumb down' data analysis, either for a non-technical audience or for experienced engineers, although it can straightforwardly produce different styles of reports for different audiences using the same input data.

Its key purpose is to make sense of the huge volume of data available, and to cut down on the time needed by engineers in the field to analyse what needs doing-and therefore increase the time and information available to them to focus on the maintenance and repairs that are needed.

It's the application of knowledge and experience, articulated in a style that an individual can understand at their own level, whether they are non-technical or an engineer with years of experience.

Oil and gas applications
In difficult-to-operate environments, for example in deepwater operations, the Arctic, or anywhere that might be considered harsh, the technology can provide operators with an option for minimum manning whilst still retaining a rapid decision-making capability.

The technology is useful in frontier exploration, particularly in the emerging nations, where data is stored in a wide range of different formats-documents, spreadsheets, CAD drawings and so on. Structured elements of these data sources can be used by NLG to produce output in articulate text with appropriately-annotated illustrations.

The technology is useful for asset integrity management. Because of the speed at which the available data can be processed and analysed, maintenance can be carried out before a situation becomes critical. The more sensors for pressure, temperature, motion and so on that a rig has, the more data there is to interpret, and the more NLG can help.

Reducing the risk of error and repair in turn leads to reduced downtime, which of course means higher productivity.

Arria NLG is a company specialising in Natural Language Generation (NLG) for mission critical industries, including oil and gas, healthcare and meterology. It has offices in London, Aberdeen, New York and Auckland.

See http://www.arria.com/

Associated Companies
» Arria Natural Language Generation

comments powered by Disqus

CREATE A FREE MEMBERSHIP

To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.