In the course of evaluating potential publishing platforms, we ascertained that a considerable number of open access platforms utilized by the humanities are based on the open-source platform OJS,[2, 3, 5, 6] developed by PKP.9Although the platform can be installed independently, we elected to rely on an experienced manager to mitigate the initial development and customization burden. In our case, the platform is hosted and maintained by Ubiquity Press,10 which has also developed its own XML parsing tool to enable readers to access articles online on the OJS interface. At the same time, in the event of a change of platform manager or a switch to an internally controlled server, the content can be moved to another instance of OJS through import and export using the native plugin, with only minimal changes.

At the time, there were no accessible word processors that could directly export in JATS format and simplify the structural formatting of the content via templates. Among the developing projects that were tested, the SciFlow Authoring platform11 had many of the desired features, despite the lack of a native JATS XML export. The software demonstrated considerable potential, particularly in regard to its straightforward management of document templates and annotation of bibliographic references. The latter is achievable both through the integrated Zotero library connection and from imported BibTeX files,12 which can be consistently updated. The formatting of the bibliography is achieved through the utilization of CSL (Citation Style Language),13 which allows the editor to modify the style during the editorial process via the selection of an alternative citation style. As a program designed specifically for scientific writing, Sciflow allows users to select structured content templates and provides guidance on the appropriate usage of paragraph styles. Furthermore, the software allows for the creation of internal references to other chapters, images, and tables in an intuitive manner, resulting in a well-organized document that is automatically populated with authors’ metadata. Consequently, starting from 2020, we elected to provide financial support for the advancement of the JATS format export and the incorporation of the article's publication metadata into the editor. This resulted in a virtuous circle, in which other institutes similarly opted to integrate SciFlow into their publication workflows. This collaborative effort culminated in the creation of the SciFlow Publishing platform, in addition to the preexisting SciFlow Authoring platform, which is now an integral part of our editorial workflow.

One of the more challenging aspects of the process is the management and extraction of bibliographic references directly from the articles where they are included in the form of unstructured formatted bibliographies. Among the numerous reference managers on the market, Citavi14 stands out due to its import functionality, which enables the loading of formatted bibliographies and their comparison with selected public catalogues. This feature facilitates the processes of content validation and data transformation.15 It is important to note, however, that Citavi also has certain limitations. The data conversion is conducted using a proprietary format that does not employ CSL, which is open-source. Moreover, the bibliographic data is stored in a proprietary format that is not readily convertible, similar to the proprietary and encrypted fields utilized within the text documents annotated via this reference manager. Consequently, we elected to utilize a Zotero Lab16 institutional license to oversee the libraries comprising the bibliographic references of our digital publications and to harmonize duplicates across articles. To ensure the stability of the identifiers, we employed the BetterBibTeX plugin to generate stable citation keys. 17

In addition, we have developed and published a bespoke CSL that is aligned with our requirements.18 It can be used directly by authors to generate bibliographic citations and bibliographies and is integrated into our customized SciFlow template for our articles.

A crucial aspect of optimizing the workflow is the collaboration with the authors. The editorial guidelines include templates for both MS Word and OO Write, which reflect the document structure required to facilitate publication. Additionally, authors are advised to utilise SciFlow from the initial drafting stage onwards. We also request a bibliographic database in a standard format, such as BibTeX or RDF, with the article references pre-inserted, thus supplementing the formatted bibliography. This serves as the basis for the editorial stage. At this time, our digital editorial program does not include open-call journals, but rather volumes of curated essay collections. Subsequently, the document is submitted via email or by sharing the SciFlow document with the editorial team. In the course of the review phase, which is conducted by the scientific committee and the peer review, a variety of strategies have been evaluated, depending on the level of curatorial engagement in managing SciFlow documents. In certain instances, reviewers were invited to provide direct commentary on online files. In other cases, papers were downloaded from SciFlow in.docx format for peer review and copyediting. Currently, the revision control, annotation, and commenting functions are not yet sufficiently developed in SciFlow to allow it to be used exclusively at a professional editorial level. Following the processing of documents and bibliographies, they can be re-imported into SciFlow, via the Import console, which is available for use with the Publishing plan. The accepted formats include.docx and LaTeX documents, with Zotero integrated references, images, captions, tables mainly preserved. During the import phase, an improved and rectified BibTeX file can also be imported, and the updated bibliographic references are annotated by simply dragging them from the entry list into the text. The pre-processed document is subsequently imported as a new SciFlow document, where author and article metadata can be added.

Given the potential for inconsistencies and typographical errors in bibliographies provided by authors, we sought tools for identifying and formatting bibliographies for the first volume's articles of our series "Hertziana Studies in Art History".19 These articles, initially intended for conventional print publication, featured a bibliography in note form that required recognition and extraction. In an attempt to automate the process, a procedure was devised that aimed to identify references within the notes and reconcile the extracted content with data from WorldCat and other sources, such as catalogues of specialised libraries (e.g., Kubikat for art and architectural history documents). [1]. The process’ instability, exacerbated by difficulties accessing catalog open data for comparative and normalization purposes, negatively impacted the workflow. Furthermore, citations subsequent to the initial one, presented in abbreviated form, were not marked or identified. In this specific instance, the construction of the bibliographic database was conducted manually during the editorial phase. The development of the tool is currently on hold,in part due to the fact that the articles submitted for the subsequent volumes had to adhere to the editorial standards for final formatted bibliography with author-date citation keys. This potentially facilitates the identification of references with existing tools such as Grobid,20 even in the absence of a bibliographic database.

JATS file export from SciFlow, particularly element-citation format that contains structured bibliographic data, features discrepancies compared to our final document versions. These include absent label with abbreviated reference citations, disorganized bibliography, incorrect document title markings, missing start and end pages for articles, exclusion of source curators, missing URIs, and lack of original date information for reprints.The manual correction of these issues is both time-consuming and susceptible to errors, particularly given the necessity of repeated copying and pasting of code snippets and manual tag insertion. Hence, with the aid of conversational AI, we've developed XSLT and Python scripts that facilitate the transformation of SciFlow's JATS file, rectifying the bibliography and integrating absent fields with the content of the BibTeX file. This ongoing refinement process hinges on the consistency of Zotero/BibTex identifiers (bar the surname-year separating dot in the BibTex version), enabling reconciliation with article content. Minor adjustments include image encoding, involving a local alternative insertion for archiving alongside the main image published from our IIIF server, and elements of the front section.21 These remain the sole manual changes performed using text editors like Visual Studio Code and Oxygen XML Editor.

A final method of manipulation of JATS XML files involves the creation of PDF files that are loaded as an alternative galley. In our case, given that we are dealing with articles that often have a preponderant iconographic content, we currently prefer to generate the PDFs in InDesign rather than as a direct transformation. InDesign allows for the loading of XML files, but specific formatting of the content is necessary to guarantee that the structure aligns with the template styles. Additionally, since the element citation field does not incorporate formatting of bibliography entries, it must be reconstructed from BibTex and CSL. Consequently, we have expanded the processing scripts to facilitate the conversion and limit the handling of complex elements to the typesetter, while still achieving a professional layout.

Given that many of the procedures mentioned previously involve using existing scripts and correcting the source code, we have decided to introduce an online interface that will act as the main workbench for all the tools, which we named PubLink. This interface allows for the application of scripts, either individually or in combination, via an online graphical interface. It comes equipped with a file manager that facilitates uploading all the elements comprising the final publication, optimizing and annotating them.

At present, the demo version is online and is primarily utilized for bibliography oversight and document preparation for OJS import.22 To manage the bibliography and extract article-related metadata for populating the OJS form, an intermediate Serial Object format has been developed to separate the JATS file elements for subsequent functionalities.

4.7.1 Bibliography Oversight. In this phase, the bibliography can be compared with either a BibTeX file or directly in the JATS XML file with the CrossRef API. Although only a portion of the references is natively assigned a Digital Object Identifier (DOI), it is becoming increasingly common for publishers to provide digital reprints of texts with DOIs. Consequently, it is possible to compare the elements with DOIs with the registered metadata, thereby identifying instances of digital reprints or missing DOIs. These can then be incorporated directly into the JATS file or exported. Furthermore, the tool is capable of analysing the JATS document in order to prepare a list of relatedIdentifiers that can be inserted in a DataCite DOI XML, thereby enhancing the metadata provided by the OJS DataCite plugin.

4.7.2 Import File Creation. As previously mentioned, importing the JATS file as a galley in OJS using the QuickSubmit function is a lengthy and tedious process. It requires both metadata entry and the need to individually upload manuscript files and all dependent elements like local images with redundant procedures. However, there is the possibility of importing complete articles with metadata and galleys from another OJS installation by using a specific “Native XML” file of the platform. This file incorporates all attachments inside as embed elements in base 64 format. Relying on existing projects, which use for example an Excel sheet,23 we have developed a transformation that can insert any missing metadata directly from the interface when starting from the intermediate Serial Object, and can add additional galleys like PDF files and cover images. Everything can be combined into a Native XML for OJS 3.3 (LTS). Upon generation, the file can be uploaded swiftly to the OJS publication platform and, after minimal checking, be published. Direct import currently creates a published file without reference issue (if not specified in the additional metadata), hence it must be unpublished to enter the correct data, a process which only takes a few minutes.

We are currently developing the integration of IIIF image annotations into articles and transforming them into XML format for import into InDesign, where PDFs are generated. Although direct PDF conversion is feasible, the humanities field frequently requires a more intricate presentation format, hence this additional step is necessary; however, this reflects a unique choice by our institute. SciFlow itself is capable of exporting the final document in PDF format with a correct pagination adhering to a custom template, making this additional phase avoidable. Additionally, if the OJS platform has not incorporated an XML to HTML parsing system, it is also possible to directly export an HTML ready for publication. If our editorial workflow would benefit from this functionality, it will be integrated into PubLink.

Our workflow, while crafted specifically for our needs, can be easily adapted for other institutes. Upon the project's completion, the source code for the PubLink platform will be made open-source,24 and its implementation through Docker containers enhances its reusability potential for other institutes. Numerous resources are available for editorial workflows, and an interface simplifying access could make digital publication as JATS XML accessible for editorial teams with lesser IT capabilities.