Kuliman "Content Profiles & linked documents"
Download as PPTX, PDF0 likes389 views
The document outlines a NISO initiative aimed at creating a standard for the interchange of academic, research, and professional content, data, and semantics on data platforms. It emphasizes the importance of treating content as data and the integration of human curation with machine learning in workflows. The CP/LD standard aims to provide flexibility and connectivity while utilizing established standards like RDF and HTML5 for effective content operations.
Kuliman "Content Profiles & linked documents"
- 1. David Kuilman, Gina Donato, Dr. Rinke Hoekstra A content standard for data-platform use cases: Content Profiles & linked documents NISO Diversity of formats February 10, 2021 11:00am Working Group initiative to create a NISO standard for the interchange of academic, research, and professional content, data, and semantics
- 2. 2 Elsevier Data Platform vision …entity-driven processes
- 3. (Early) access and visibility Expedite shapes Lineage Provenance Policy / license Priority of content and authorship Content is data Content and data operate seamlessly Content structure follows document entity structure Rich HTML5 literals for UI/UX use cases Role based processing Content typology Granular Context-based using process and purpose intelligence Content is shared All content can be leveraged throughout the platform by all contributor/consumer roles using a common vocabulary Zero organisational boundaries Policies for compliance Continuous flow and hydration Partial and complete resources Extensible types and enrichments Optimisation of formats Machine learning Human interaction Agile, extensible and resilient Fast services development Nimble models Extensible models Arbitrary content (types) Service level agreement Handle exception flows gracefully and informed Business requirement: from a content perspective
- 4. Anatomy of content entity processes on a data platform Source Data Harvesting Normalisation Extraction matching Linking Curation Publishing … entity driven workflow Classic document driven workflow… manuscript Internal format copyedit Mastercopy Product mappings mappings
- 5. The Content Profiles & Linked Document standard (CP/LD) is the result of adopting content platform principles to provide the flexibility, extensibility and connectivity required on a data platform for academic, research and professional content Lets consider a few critical design considerations first… Pipeline to cyclic Human-in-the-loop Merging data entities and content entities on demand
- 6. Sourcing Harvesting Normalizing Extraction Matching linking Publishing Key concept: think cyclic, not linear… Sourcing Harvesting Normalizing Extraction Matching linking Publishing Sourcing Harvesting Normalizing Extraction Matching linking Publishing Sourcing Harvesting Normalizing Extraction Matching linking Publishing Sourcing Harvesting Normalisation Extraction matching Linking Curation Publishing … in parallel workflows … author … review … approve … connect … edit … recommend … annotate … Human-in-the-loop
- 7. Key concept: think human-in-the-loop and machine learning Sourcing Harvesting Normalizing Extraction Matching linking Publishing Gold set Test sets Human curation within content centric workflows Human curation within Machine Learning Contributor Consumer Continuous improvement Content operations Platform operations Continuous deployment Model operations Content artefacts Enhanced Content artefacts Human supervised Content usage metrics
- 8. The CP/LD standard uses established standards to create the format framework that supports data platform content operations without compromise Linked data and HTML5 unite syntax, structure and semantics needed on the platform
- 9. HTML5 JSON-LD + Structured narrative Semantic data layer XHTML dialect Linked Data Usage standard and guidelines Independent of any particular use case Content Profile standard & Linked Document XML Schema RDF Schema SHACL XML Schema RDF: Discovery XML: consistency JSON: messaging JSON-LD: knowledge infusion HTML5: representation Business roles
- 10. This is a part of text that has a specific style (italic) This is a paragraph This paragraph is the abstract of the paper This paragraph is the title of the paper This is author Alba Grifoni This is a citation of another paper This is a result reported on in this paper This is a mention of the “COVID-19” concept This is a mention of the “SARS-CoV2” concept This states that “SARS-CoV2” reactive “CD4+ T-cells” exist in ~40%- 60% of unexposed individuals, suggesting cross-reactive T-cell recognition with “common cold” doi:10.1126/sciimunol.aan5393 “55425663600” hgraph:id-88f9e4ca-c776-3380-933b-f1218c4ef1fd (COVID-19) hgraph:id-2ab6cd87-e543-3229-85ff-c862a90f415c (SARS-CoV2) hgraph:id-88f9e4ca-c776-3380-933b-f1218c4ef1fd (T-CD4+) hgraph:id-2ab6cd87-e543-3229-85ff-c862a90f415c (SARS-CoV2) hgraph:id-a28e7725-1919-34f0-a648-45721d8bd6a2 (common cold) reactive to reactive to The anatomy of a Linked Document
- 11. service service service service service service service service service service assertions documents resources Aggregations products Content Topics blueprint for data platform Bespoke normalizers Linked Data processors Query harversting Harvested manuscript Normalized document Enriched article A finished article
- 12. Article Author Document Document Document Author Document Article Document Author attributes Manuscript Conclusion Abstract Author String Author String Activating the platform: listen and merge application An author manuscript Author mention Author as Person Entity Author as Entity and representation Conclusion Abstract service service service merge
- 13. Activating the platform: merge topics and create a product view After merging the topics, the finished view offers: • A manuscript becomes an Document • the position of an abstract and a conclusion • An person has been identified as author • The author string has been identified within the document. • The author has entity attributes • The document assembly is a scientific article of type ‘Finished’ because it satisfies the above criteria merge Article Author Author attributes Abstract Author String Conclusion Outside document Inside document HTML5 vocabulary JSON-LD predicates Relationships legend A finished article
- 14. Key takeaways • Content is data; treat it as data not as documents • Normalization is great divider from files to entities, items and assertions • Entity-designed data and Author-designed data become blended • Machine learner and researcher forge alliance On standards & formats… • RDF and XML schema technology (remain) backbone for information modelling • JSON, JSON-LD and HTML5 serialisations dominant for content standards Working Group initiative to create a NISO standard for the interchange of academic, research, and professional content, data, and semantics Further information:
Editor's Notes
- #11: XML DTD 5.6 (OPS), XOCS… Common Index Profile (CIP) -> structure & metadata NLP: CM2, FPE, Leadmine, MedScan, Termite (SciBite) … Linking: Parity, FPE, …