You can think of an #Ontology as being like a #Map!
By way of #analogy if you think about the London #AtoZmap, and the #TubeMap, they are used for different #purposes, different #tasks, and they #represent the #world in slightly different ways. Sometimes you can find ways in which they #align#conceptually, and other times they don’t; and if you were to build some #software which sought to leverage these #representations they would work in different ways. This is a good analogy to what an Ontology is. Ontologies are powerful ways of organising representations that can be leveraged by #AIsystems, typically #logic based reasoning systems; but ontologies can also be used to create #features that can be used to support #labelled#machinelearning as opposed to the current fashion for #unlabelledlearning systems, such as ChatGPT. In some respects this #distinction around labelling is something which easily #obscures the fact that if you are #training a system on so called unlabelled text it already inherently has a structure in the choice of the #alphabet that you use, and the combinations of #letters into #words, and words into #sentences, that inherently represent a #language – such as #English or #Portuguese that can be identified by common #patternsin the so called unlabelled data that you feed into your system. Which also leads to an #implicit ontology.
So, that’s a bunch of thoughts for you around #representations, #semanticsand #artificialintelligence on #JabeOnAI
