As of this month, American Chemical Society (ACS) is using Writefull’s Metadata API to automate aspects of manuscript review. Writefull’s Metadata API leverages state-of-the-art NLP to parse documents and list their components in a structured response file. This service facilitates the (automated) review of manuscripts, offering a scalable, time-saving solution to publishers.

Why Writefull’s Metadata API

Publishers usually review manuscripts to ensure all elements are complete and consistent: think of key sections, affiliations, and author names. Manually reviewing these elements can be time-intensive and hard to scale. Writefull’s Metadata API helps by automatically extracting manuscript elements, for example, for structural checks or for quality assessment after editing.

ACS’s use case

ACS and Writefull have been collaborating for nearly two years, defining and testing ways to automate parts of the publishing pipeline. While the main focus has been the use of AI for language-related tasks, ACS was also looking to speed up the internal manuscript checks that are part of their quality control program.

ACS assesses the quality and accuracy of their copyedited manuscripts using an elaborate scorecard process. Included in the scorecards are aspects such as whether author names and affiliations are present and correct.

ACS now uses Writefull’s Metadata API to facilitate and partly automate this task. The metadata from the API response is directly compared to elements in the manuscript XML after editing. Where items match, the scorecard is automatically populated. Where items do not match, a report is generated that highlights the mismatches for manual review by ACS’s quality control staff, reducing both the number of scorecard items requiring manual review and the time to review them.

How Writefull’s Metadata API works

Writefull’s Metadata API uses advanced NLP approaches to extract elements from manuscript files that were created without a template (i.e., unstructured manuscripts).  In such cases, simple, hard-coded rules cannot be applied because the content in these elements can vary greatly. For example, consider addresses or author names, which do not follow predictable patterns and cannot be accurately structured using rules alone (e.g., compare the ability to identify the surname among John Doe, Jan-Paul van der Veen, and Paola Baron Toaldo).

Writefull’s Metadata API leverages thoroughly trained and vetted AI models to tackle this problem. These models turn unstructured and often noisy data into a structured object with fine-grained properties. The API parses DOC/X and PDF files and provides a structured JSON response that lists the elements. This JSON file can be read to quickly spot gaps and errors, or it can be integrated by the publisher to facilitate or automatize their in-house manuscript checks. In the case of ACS, it is used as part of their quality control process after editing. Other publishers, such as Hindawi, use it for structural checks.

More information

Would you like to learn more or discuss how Writefull may fit your publishing workflow? Please contact us at