Copyediting of scholarly materials is a complicated topic. Authors differ significantly in being able to clearly and accurately communicate their work in writing. These differences can be based both on an author’s native language as well as the amount of focus that has been placed on writing skills in their educational background.
As a result, scholarly publishers place a lot of time, focus, and money on editorial processes that are meant to standardize output content at a high level of language quality and industry styles. In most cases, this means that every article, regardless of a demonstrated need for editing, goes through a full copyediting process.
The thing is copyeditors edit copy, even when it may not need much editing. And further, what we’ve found is that many authors are pretty good writers. Our research shows that the #1 complaint most authors have with the scholarly publishing process is their content being changed by a copyeditor who is not making the material any better, except doing few changes.
At Katalyst, using artificial intelligence (AI) and machine learning, we’ve found a better way of managing such things. Our production teams are now using an innovative new tool called CC3SM (Contextualized Copyediting – 3 levels) to separate “well-written” content from content that is, “not-so-well-written.” Evidence from live production now shows that this workflow reduces editorial costs and shortens overall schedules by 50% or more.
CC3SM works by first structuring content to house or industry style (pre-editing, which includes conversion to XML, application of style guidelines, reference structuring, correction of typos and US-UK English usage, along with other automated tasks), and then passing it through a proprietary AI-based natural language processor that we’ve developed, which scores the content in multiple areas of English language use. Content that scores above a certain threshold is categorized as Level A and sent through the production process with little or no editorial intervention. Content scored at Levels B and C are assigned an appropriate level of editing. Our research tells us that roughly 35% of incoming material falls into Level A, 45% Level B, and 20% Level C. This additional level of intelligence in the editorial workflow passes a smaller percentage of articles through the traditional copyediting process, speeding time to publication and reducing copyediting cost.
Katalyst’s Nova Techset presented a joint case study on CC3SM at the 2018 annual meeting of the Society for Scholarly Publishing (SSP) held in Chicago May 30 – June 1, 2018. Stewart Gardiner, head of the Taylor & Francis global journals program commented, “Our perspective on this is simple: If an article needs work, we should be copyediting it. If an article doesn’t need work, we should leave it as it is. The results have been very positive. We regularly survey authors and editors, and our data shows a high level of confidence in our editorial work using Contextual Copyediting.”