7 November, 2017
Technology has invaded all areas of our lives, and the editing process is no exception. Editing and proofreading software like PerfectIt is now widely available, and is improving year on year. So, if machines are getting better at doing the copy-editor’s job, will that mean we don’t need human editors anymore? It may sound like an inevitability, but that misunderstands both the value of a human editor and the capabilities of artificial intelligence. Here’s why you will never be able to rely on software to take the place of an editor.
Artificial intelligence (AI) is not one technology, but a collection of technologies such as advanced analytics, expert systems, neural networks and machine learning. Some form of AI is being used in a wide range of applications, most successfully those that are rules-based: driverless cars, targeted advertising, digital assistants, photo tagging, strategic games, medical diagnosis, and interpreting complex data. These are all very powerful tools, yet they all lack one thing (most!) humans have in abundance: common sense. The few machines that get close to passing the ‘Turing Test’ have done so through trickery and deception – the electronic equivalent of the conjurer’s sleight-of-hand.
The editing process is a three-way dialogue between author, publisher and editor, undertaken within an editorial context that changes with every single publication. The editor will approach the document with a lifetime of language experience and a knowledge base of facts, perceptions and memories culled from archaeology to zoomorphism, all crucially overlaid with empathy and sensitivity to the author’s written word.
The author’s voice cannot easily be simulated. Important elements of the author’s style – a key attraction of any book – may be eliminated or changed if left solely to an artificial editor. For example, euphemisms can change a word’s impact, create a particular atmosphere, or signal political or social views. In the first paragraph of this article the word ‘invaded’ was used. What difference to the tone would ‘permeate’ or ‘infest’ have made? Different words carry different connotations, which can be far more powerful than any dictionary definition, and which arise over time from context, history and life experience.
“Grammar is complex. Writers may even want to break rules sometimes, for example, to accurately represent what a source said or to create natural sounding dialogue. Software may help, but editing is about much more than that.”— Mignon Fogarty, Grammar Girl
All these subtleties are easily recognized and understood by the human editor. By contrast, a piece of software can ‘know’ the definition of a word, but can’t understand the meaning of that word within the larger context. It can’t identify the emotional response that each and every word elicits.
Why is language so difficult for machines? According to Noam Chomsky, the linguist, philosopher and creator of the Universal Language Theory, language is a process of free creation. Almost every sentence we speak is a brand-new combination of words that may never have appeared before. To be able to do that, the brain must create infinite combinations out of finite lists of words. The rules for using these lists may be fixed, albeit sometimes ambiguous and fluid, but the manner in which the rules are used to generate language is free and infinitely varied.
The language learning process itself gives a clue as to how difficult it is for a machine to simulate human language. Children learn language by accumulating vocabulary – 10–15 new word meanings a day – but research indicates that most words aren’t directly taught. They are gleaned from the child’s experience and the context in which the words are used. There is no equivalent process for AI. And if that wasn’t enough, remember that language itself is constantly evolving so quickly that dictionaries can’t keep up with the linguistic innovations. So can you imagine what an editing software package would do to Shakespeare?
This is not to say that software has no place in the editing process. The latest generation of editing software helps copy-editors and proofreaders with the mundane and (let’s face it) mind-numbing part of the job. Macros and algorithms are available for checking grammar and spelling, summarizing documents, eliminating inconsistencies or finding duplications.
PerfectIt is an outstanding tool for helping copy-editors speed up the mechanical side of their job, improving their productivity and allowing them to get to the more creative side of the process more quickly. As well as checking style, consistency, and aspects of punctuation and spelling, PerfectIt can be customized. Copy-editors can establish rules in PerfectIt to find and replace specific words/phrases/punctuation, or incorporate their client’s house style into its procedures. But when PerfectIt has finished, the real and most valuable part of the editor’s job starts.
Let’s not confuse aids and tools with artificial intelligence. Vendors like to brand their software ‘AI Inside’ – but that doesn’t mean you should believe the hype. Intelligent Editing, the authors of PerfectIt, take a different approach. Their mission statement says, ‘humans make the best editing decisions and they always will. We build technology to help them edit faster and better.’
Put simply, true AI software for editing does not yet exist, and there are two major reasons why it will never replace the human editor. Firstly, editing is about more than correcting text. It’s about understanding and manipulating linguistic shade, colour and nuance; software works in black and white. And secondly, and probably more importantly, language is just so insanely complex!
Optimists believe AI will lead us into a golden age where disease and poverty are eradicated. Doomsayers foresee machines with a will of their own in conflict with humanity. But copy-editors and proofreaders shouldn’t be alarmed by the rise of AI. The newly released Blade Runner 2049 has lots of characters with artificial intelligence. Not one of them is an editor!