Document Co-authoring Enhanced by Structured Knowledge

Semantic Web Interest Group,


Introduction

Hello. I am excited to share some ideas — brainstorming really — about future document-authoring scenarios enhanced by structured knowledge extracted from natural-language documents using artificial-intelligence systems.

These ideas were recently inspired by an open-source project and corresponding prototype created by another team. Hyperlinks to it are available in the postscript [1][2].


Next-generation Project Templates

Expanding on the concept of document templates available today in word-processing applications and websites, in the not-too-distant future, users might be able to choose which kind of document-related project that they intended to do, selecting and then making use of a next-generation project template.

These next-generation project templates could additionally provide sets of guidelines, rules, and requirements (see: [1][2]) for interoperating artificial-intelligence assistants to utilize when providing assistance with respect to documents. These guidelines, rules, and requirements could either be embedded in or hyperlinked to from these next-generation project templates.

These features could enhance man-machine document co-authoring in a number of exciting, project-specific ways.


Extracting Structured Knowledge from Natural-language Documents

While foundation models don’t need to extract structured knowledge from documents to perform many helpful tasks, well-formed documents, in conformance with project-specific guidelines, rules, and requirements, could, in theory, more readily have structured knowledge extracted from them. Project templates could be used to provide contextual cues, schemas or ontologies, questions for, prompts for, interactions or scripts for, foundation-model-based artificial-intelligence systems.

The structured knowledge extracted from natural-language documents could be validated and processed using those schemas or ontologies specified in project templates. These schemas or ontologies could either be embedded in or hyperlinked to from the project templates.

With extracted structured knowledge available to artificial-intelligence systems alongside natural-language documents, more varieties of processing could be performed to provide individuals and teams with more features.


Foundation Models and Knowledge-based Artificial Intelligence Working Together to Provide More Features for Document Authors

The results of processing and reasoning about natural-language documents and their extracted structured knowledge, using a wide variety of technologies and tools, could be routed to users' AI assistants for these to provide yet more helpful features and assistance for individuals and teams.


User Experience Concepts

Individual authors and teams of authors could make use of a tabs-based user interface or menuing system to switch between views of their natural-language documents and visualizations of the structured knowledge extracted at that point, these displayed or visualized utilizing, perhaps, project-specific schemas or ontologies or other related resources.

As one can observe in the shared prototype [1][2], artificial-intelligence systems could utilize visual highlighting to indicate areas in documents where any guidelines, rules, or requirements were applicable to the natural-language document.

Artificial-intelligence systems could additionally open or update issues about co-authored documents in shared repositories or other issues-based collaboration spaces.


Conclusion

Perhaps man-machine document co-authoring scenarios can, in a sense, be viewed as a form of advanced, interactive data entry resulting in both natural-language and structured-knowledge artifacts. People would, then, be able to publish their documents and resultant structured-knowledge artifacts, e.g., XML-based or Semantic Web resources, alongside one another.

Thank you. I hope that these ideas were of some interest to you. Any thoughts on these ideas?


Best regards,
Adam Sobieski

[1] https://github.com/wikius/omnipedia
[2] https://omnipedia-client.pages.dev/