Friday 3rd April 2020

AI in Aerospace Technical Writing - Introduction

Will the laborious technical publication sector of the aerospace industry be conquered by AI? How close or further are we from such a revolution? How far it might go and what all are the potential impacts? Let us analyze various aerospace industry facts, spots on timeline, and technological aspects and try to simulate a probable future scenario.

Who is going to do it better?

Unlike any other product supporting documentation, aircraft maintenance documents are the most complex and extensive set of written information which are being produced for the end-users since the age of  Wright brothers.    

Before many blue moons, before the age of the internet and electronic publications, the document racks of Maintenance, Repair, and Overhaul Organizations (MROs) were more vast and complex than an aircraft itself. It was a herculean task for the maintenance and operations crew to manage different revisions of aircraft manuals aka publications. Even if two aircrafts are of the same model, there are many intrinsic uniqueness to each serial number. The emergence of electronic platforms eased the storage of these publications and made it less tedious to navigate through them.

As SAFETY is the keystone of the entire aerospace industry, there is no room for error or omission of any technical information which the end-user refers to. This is the reason why the aircraft manufacturers spend millions to generate, maintain, and update the aircraft publications. Thousands of aerospace technical writers with engineering background, all around the globe are working to fulfill this humongous task. They work round the clock to keep up with the engineering design upgrades of existing aircraft models as well as newly engineered models.

A steady hype of "Machine learning" shown in Google trends in the recent years.

The possibilities of machine learning have exploded in the recent years, finding its way to different streams of finance, transportation, healthcare, retail industry, agriculture and many others. In contrast to the previously experimented hand coded rule based language processing systems, evaluation of machine learning algorithms based on deep neural networks certainly has a huge potential in aerospace technical writing. The commercial release of "Boeing Simplified English Checker" in 1990 is an example of the hand coded system. Even though they hand coded ASD-STE 100 rules with a vocabulary system and profiler to complement, it is nowhere near to the potential language processing which we can do by applying machine learning algorithms. Since the late 1990's, there have even been university level research papers published which had the potential to revolutionize aerospace technical writing.

The obvious question here is then why the aircraft manufacturers or even their software solution partners have not embraced the full possibilities of AI in technical writing? Let me be the devil's advocate here (as some of us believe, AI is going to be the devil who will take over the world) and try to answer it with a series of other questions.

Do we really trust the computers to write the aircraft maintenance procedures by itself? Even if we make up our minds to trust the computers (after all, they do the auto-land safely. lets cut them a little slack here), how will the computer analyze the entire set of engineering documents and convert it to relevant data in various schemas for the end-users? Alright! Engineering documents are already in digital format decades and already in the big belly of computers. Still will it ever be able to understand the S1000D standards and follow the company specific and program specific style guides, guidelines, and templates? Without machine learning itself, it can be achieved by a great team of programmers and industry specialists. After all, all standards and guidelines are rules. Computers love to follow the rules and do the job. My fellow programmers would agree with me in a heartbeat. A self learning machine can not only follow the rules but identify new rules and adapt. Its learning capabilities can analyze, identify, and apply new patterns from multiple sources simultaneously which a technical writer would take many years to gain some level of competency. Now the million dollar question; Even if computers can do all these, do we really want to let all our aerospace technical writers and subsidiary professionals go jobless?

To find the answers for all these questions and dive into the possibilities and limitations of machine learning in Aerospace, subscribe to my profile and wait for the series of blogs coming up. I will take aboard every aerospace technical writer who worries about their future in their profession. Your subscription to my profile will be great encouragement to make this a successful and profound series of blogs. I will share information on not only the impact of AI in technical publications but also different streams of the aerospace sector.  I will review and share all the technological advancements from the common grounds of AI and Aerospace. As Gorden Gekko says in the movie Wall Street (1987) - "The most valuable commodity I know of is Information".