The Future of Knowledge Organisation: Semantic Web and Beyond

The web you use every day was built for humans to read, not for machines to understand. A search engine can find a page that contains the words “Tagore” and “Nobel,” but it does not truly know that Rabindranath Tagore won the Nobel Prize in Literature in 1913. That gap between text and meaning is exactly what the next generation of knowledge organisation is trying to close. Libraries, archives, and information centres have spent over a century building classification schemes, thesauri, and subject headings to organise knowledge. The question now is how these traditional systems survive and thrive on a web that is increasingly read by software, search bots, and artificial intelligence. This is where the Semantic Web enters the story.

What the semantic web actually means 🔗

The Semantic Web is not a separate internet. It is an extension of the web we already use, designed to give data well-defined meaning so that computers and people can work together more effectively. The idea was introduced in 2001 by Tim Berners-Lee, the inventor of the web, along with James Hendler and Ora Lassila. Their vision was a web of linked, machine-readable data where intelligent agents could carry out tasks on a user’s behalf.

To make this work, information has to be structured. On the traditional web, a page about a book is just a block of text. On the Semantic Web, that same information is broken into small, logical statements that a machine can process. Each statement follows a simple pattern: a subject, a predicate, and an object. For example, “Premchand (subject) wrote (predicate) Godaan (object).” Thousands of such statements, linked together, form a web of meaning rather than a web of documents.

From documents to linked data 🔗

This shift has a direct impact on knowledge organisation systems, often shortened to KOS. A KOS is any tool that organises knowledge: a classification scheme like Dewey Decimal, a thesaurus, a taxonomy, a list of subject headings, or an ontology. For decades these systems lived inside library catalogues and printed volumes. The Semantic Web invites them onto the open web as structured, shareable data that any system can reuse. Knowledge graphs, which power features in modern search engines, are a practical result of this idea. They link entities such as people, places, and works so that a search for one concept can surface related ones, turning isolated facts into an interconnected network where relationships reveal new patterns.

The challenges holding KOS development back 🔗

Moving traditional organisation systems onto the web is not simply a matter of uploading old files. Several real problems stand in the way, and understanding them is essential for anyone studying information science today.

Interoperability and semantic conflicts 🔗

The biggest hurdle is interoperability, which means getting different systems to understand each other. A medical library, a national archive, and a university repository may each describe the same concept using different vocabularies. When their collections need to be searched together, these mismatches create confusion. Mapping between different KOS, both within one language and across languages, remains a pressing concern if heterogeneous collections of resources are to be integrated. This is why semantic interoperability is treated as a core goal for organisations that provide access to information across different contexts. Without it, structured data stays trapped in separate silos.

Data structuring and maintenance 🔗

Structuring data correctly is demanding work. A KOS published on the web is rarely finished. Vocabularies are updated as new concepts emerge, and every new version can break the links that other datasets have built to the older version. Updating a controlled vocabulary is a knowledge-intensive task that requires careful deliberation, and when a new version is released, users who linked their data to the previous one may find their links are no longer valid. These version chains create ramifications for both publishers and the people who depend on the vocabulary. Maintaining quality, consistency, and scalability over time is one of the hardest parts of implementing knowledge graphs in real library settings.

User-centric design 🔗

Finally, these systems must serve actual people, not just machines. A perfectly logical ontology is useless if researchers and students cannot navigate it. User-centric design focuses on how someone actually searches, browses, and discovers information. KOS support this by enabling query expansion, where the system uses broader and related terms to help users execute precise searches and discover more relevant content. The challenge is designing interfaces that hide the technical complexity while still delivering its benefits.

The role of W3C standards: RDF and OWL 🔗

None of this would be possible without shared rules. The body responsible for these rules is the World Wide Web Consortium, or W3C, the same organisation that maintains the core standards of the web. For the Semantic Web, the W3C has developed a family of standards that allow data and meaning to travel across systems.

RDF: the foundation 🔗

The Resource Description Framework, or RDF, is the base layer. It introduced the graph-based model where knowledge is expressed as triples of subject, predicate, and object. Because every piece of information is broken into these atomic statements, RDF allows data from completely different sources to be combined into one connected graph. It is the common grammar that lets a record from one database join smoothly with a record from another, making it the foundation for representing semantic graphs that enable interoperability, knowledge discovery, and automated reasoning.

OWL: adding logic and reasoning 🔗

The Web Ontology Language, known as OWL, builds on top of RDF and adds far more expressive power. OWL allows you to define classes, properties, and the logical relationships between them. Crucially, it supports inference: a reasoning engine can work out new facts that were never stated explicitly. If the system knows that all novels are books, and that Godaan is a novel, it can conclude that Godaan is a book without being told directly. OWL is a more complex language that allows richer inferences and can express constraints, which is useful for checking the consistency of a model and catching errors in how knowledge is structured. Together, RDF and OWL turn raw data into knowledge a machine can reason about.

Future trends shaping knowledge organisation 🔗

The future of knowledge organisation is being shaped by three forces pulling in interesting directions: the energy of ordinary users, the precision of formal standards, and the power of artificial intelligence.

Social tagging and folksonomies 🔗

One major trend is social tagging, where users freely attach their own labels to content. The collection of these user-generated tags is called a folksonomy. Unlike a controlled vocabulary built by experts, a folksonomy grows organically from the community. Its strength is flexibility and the way it captures language as people actually use it. Its weakness is the lack of standardised rules, which leads to inconsistent tagging and quality-control problems with user-generated metadata. The exciting research question is whether the two approaches can be merged. Information scientists ask whether folksonomic tag clouds can be combined with controlled vocabulary systems to create hybrid systems that take the best of both worlds.

The rise of the SKOS framework 🔗

Bridging the old and the new is exactly what the Simple Knowledge Organization System, or SKOS, was built to do. SKOS is a W3C recommendation, finalised in 2009, that provides a standard model for expressing thesauri, classification schemes, taxonomies, and subject heading lists in RDF. Its great achievement is offering a low-cost migration path for porting existing knowledge organisation systems onto Semantic Web technology. This means decades of carefully curated library vocabularies do not have to be thrown away. They can be published as linked data and reused across the web. SKOS deliberately sits in a middle position between informal social tagging and the strict formal logic of OWL, making the accumulated wisdom of library science usable in a machine-readable world.

Artificial intelligence and knowledge graphs 🔗

The most powerful current trend is the integration of artificial intelligence. AI and machine learning are increasingly used to build, populate, and maintain knowledge graphs in digital libraries, supporting more intelligent and user-centric information services. Large language models add a new dimension by allowing systems to combine structure with statistical meaning. A growing approach stores both the explicit relationships of a knowledge graph and the learned patterns of text embeddings together, aiming to unite structured search and semantic meaning into a single system. This blends symbolic, rule-based reasoning, which is transparent and explainable, with the flexible pattern recognition of modern AI. As these technologies mature, the promise is greater interoperability and far more personalised discovery, moving from generic results toward understanding individual user needs at a much finer level of detail.

Why this matters for the information profession 🔗

The skills librarians and information professionals have always practised, organising, describing, and connecting knowledge, are not becoming obsolete. They are becoming the foundation of the intelligent web. Classification and controlled vocabularies are precisely the structured, trustworthy knowledge that AI systems need to stay accurate. The professional who understands both a traditional thesaurus and an RDF triple is exactly the kind of expert the future demands. The web is shifting from a pile of documents to a connected map of meaning, and knowledge organisation is the discipline drawing that map.

What do you think? Could a hybrid system that blends expert-built controlled vocabularies with community-driven social tagging finally solve the consistency problem, or will the two approaches always pull against each other? And as AI takes on more of the work of organising knowledge, what should remain firmly in human hands?

References

1. https://cacm.acm.org/research/a-review-of-the-semantic-web-field/
2. https://arxiv.org/pdf/2502.15689
3. https://www.tandfonline.com/doi/full/10.1080/13614560600856433
4. https://www.academia.edu/Documents/in/Knowledge_Organization_Systems?after=119932464
5. https://direct.mit.edu/qss/article/doi/10.1162/qss_a_00363/128602/A-survey-of-knowledge-organization-systems-of
6. https://graph.build/resources/semantic-graphs
7. https://www.w3.org/2007/03/VLDB/
8. https://oercommons.org/courseware/lesson/123503/overview
9. https://www.w3.org/TR/skos-reference/
10. https://link.springer.com/chapter/10.1007/978-3-031-69216-1_9
11. https://neo4j.com/blog/developer/knowledge-graph-structured-semantic-search/
12. https://www.dataversity.net/articles/semantic-web-and-ai-empowering-knowledge-graphs-for-smarter-applications/