It’s about aboutness: Semantics Conference, Karlsruhe, 09-12 September 2019

The semantics of the semantic web

As a former linguistics student, I am always perturbed when people try to dismiss an argument by saying ‘it’s just semantics’ – as if the way in which humans make sense of the world through language is somehow of minor importance. The concept of the ‘Semantic Web’ is familiar to many, but still not widely understood: as Michael Ushold asked in his 2003 article, ‘Where are the semantics in the semantic web?’. Of course, at the time Ushold was writing, the Semantic Web was still a largely theoretical concept: it is fair to say that only recently have developments in machine learning and artificial intelligence made it possible for applications using these principles to be realised. Indeed, the title of the public meetup held during the conference was ‘How knowledge graphs provided Semantic Web with a second spring’ – a theme much in evidence throughout. 

Things, not strings

An excellent overview of the principles of knowledge graphs was given by Heather Hedden, the ‘Accidental Taxonomist’, and Andreas Blumauer of The Semantic Web Company, in the pre-conference workshop entitled ‘Fast Track to Knowledge Graphs and Semantic AI’. Covering taxonomies, ontologies, thesauri, knowledge models and the fundamentals of linked data, it provided a solid foundation for the more in-depth conference sessions that followed. The standard of talks was extremely high, with a good blend of theory and practice. I was particularly interested in the sessions on text analysis, natural language processing and the related field of natural language understanding, an area with clear relevance to both machine translation and the development of digital assistants. The topic of understanding was raised again in one of the standout sessions, Valentina Presutti’s keynote talk on ‘Looking for Common Sense in the Semantic Web’. Common sense is a highly contested topic, but one which is by definition rooted in human social understanding. Linked data and knowledge graphs can get us closer to retrieving hidden information and uncovering ‘shadow concepts’ which are implied, though not explicitly named, from texts, but cannot replicate the associations produced by human thought processes which are based on real-world knowledge rather than logic. As another speaker mentioned in a different context, the ‘sameness’ of things is a semantic minefield. 

Artificial intelligence or assistive intelligence?

As we move into an era where AI becomes ‘assisted intelligence’ or even ‘assistive intelligence’, not only will our relationships with machines – such as virtual assistants and domestic robots – change, but so will our relationships with our fellow humans. What happens to the vision of the ‘web of trust’ when we are faced with Twitter bots and deep fakes? If we are to be true to the principles of the Semantic Web, we need to remember that linked data, knowledge graphs and other semantic technologies are primarily about relationships and the real-life connections not only between concepts, but, crucially, between people.