Right-sizing language editing using AI: a case study in improving publication speed and author satisfaction
The application of artificial intelligence to areas of traditional craft skill in the academic publishing industry creates an opportunity to reshape the knowledge supply chain and improve the velocity of scientific communication. This session explores the lessons learned from a recent pilot program at Taylor & Francis journals to apply natural language processing algorithms to accepted articles in order to intelligently stratify incoming articles to differing levels of editorial treatment based on the language quality of the input material. We share data derived from Taylor & Francis’s author and editor surveys to illustrate the surprisingly inverse correlation between the level of human intervention and corresponding perception of copyediting quality. The session provides a technical demonstration on how the technology is deployed in practice in a prepress workflow operating at a large scale of article throughput. It concludes with an examination of the insights derived from the pilot program’s data on the distribution of article language quality and intervention requirements for typical journals, and the before and after effect on average article transit speeds and overall author satisfaction. Further applications for AI in the peer review and production process are posited. Overall the findings of the pilot demonstrate that the application of AI to automating decision- making in prepress workflows produces encouraging results and deserves further exploration.