Data First
Another very thoughtful posting by Stefano Mazzocchi. This issue of pre-imposing structure on data reminds me of the John Tukey’s pioneering work on EDA (exploratory data analysis) in the field of statistics. Up to that point, most statistical analysis consisted of “confirmatory” (structure imposed) methods rather than the various “exploratory” methods (including data visualization) introduced by Tukey. A role of exploratory data analysis is to suggest patterns in the data for further inquiry or possible modeling which may then rely upon confirmatory methods for validation.
Posted by Mark at
Yes very interesting article and thoughtful. Indeed. I have again an in-between position, and maybe it’s what Stefano was saying actually in the conclusion of his post.
It’s not that much data first/structure first, because we would fall into the same characterization of the chicken-egg problem. I would say it’s both of the things happening at the same time.
We learn from new data (because we have structure to digest this experience) then we create new structures paradigms with these data and we are able to collect more data, in an infinite loop. The language is all about that. We have a structure (grammar) which helps to articulate data (words, facts, etc.) that will help to continue the process. We could say “Hey the meaning is into the content”. Well yes, but without the grammar to understand it, you are blocked, you can’t do reasoning. At the same time, you could have grammar but without data, you will not go very far too.
What I like very much about RDF is that I have the right to be messy in collecting my data compared to XML. :) XML is very rigid to write with RDF I draw, I create relationships between things exactly like I see the world around me. It doesn’t mean I reject XML, it’s very useful in many circumstances, but RDF offers so much freedom to collect data and build step by step my knowledge. I don’t need to know the structure beforehand, because I can create it and plug it to others easily. It’s very biological.
Posted by karl Dubost atDonnées ou Structures
Dans un récent billet, Stefano a illustré deux modèles conceptuelles de traitement de l’information : Données et structures. Il traite les deux systèmes en posant la question de la priorité et de l’efficacité. Apprend-on des données brutes et...Excerpt from Karl & Cow - Le carnet Web at
Game time started
Short but powerful piece from Stefano Mazzocchi entitled Data First vs. Structure First . I normally disagree with a fair proportion of Stefano’s material, and naturally disagree with anything presented as a general rule (especially ones Sam links...Excerpt from Planet RDF at