Hacker News new | past | comments | ask | show | jobs | submit login

Having worked in this field for a bit, this uncovers an even more fundamental flaw: The idea that we can have a single static ontology.



Domain driven design is well aware that is not feasible to have a single schema for everything, they use bounded contexts. Is there something similar for the semantic web?


In the Semantic Web, things like ontologies and namespaces play a role similar to bounded contexts in DDD. There’s no exact equivalent, but these tools help different schemas coexist and work together


Isn't that the point of RDF / Owl etc.?


The point of RDF is to raise the problem to the greatest common divisor level of complexity - a hypermedia graph of arbitrary predicates about arbitrary objects.

OWL is a modeling language to describe ontologies, e.g. some constraints people have agreed to follow about how to structure the information they publish in graphs. It can also be considered an advanced schema language.

The idea of a bounded context in DDD is that it is not a good use of time (or indeed may not be feasible at all) to get a single ontology for an entire domain, so different subdomains may be unified by some concepts but have differing or overlapping concepts that they use internally. Two contexts know they are talking about a product called "New Shimmer", even if understands it as a floor wax and the other uses it as a dessert topping.

The two pillars of the semantic web are public data and machine understanding, which IMHO pushes strongly toward the (often unachievable) goal of a single kitchen-sink schema.


Mostly, the problems of a semantic web are covered in the history of Cyc[1].

When I started to use LLMs I thought that was the missing link to convert content to semantic representations, even taking into account the errors/hallucinations within them.

[1] https://en.wikipedia.org/wiki/Cyc


There is also the problem that structure doesn't guarantee meaning.


What kind of work do you do?


various. Notable I, some years ago, had a project that considered automatic consolidation of ontologies based on meta-ontologies and heuristics.

The idea being that everyone have their own ontology for the data they release and the system would make a consolidated ontology that could be used to automatic integration of data from different datasources.

regardless, that project did not get traction, so now it sits.


Thanks! Is it possible for you to share more details about the domain and problem?




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: