Talk:Metadata

From Pelennor

Metadata patents

• The gruesome visage of intellectual property law strikes against innovation and progress yet again (maybe). It appears that Apple filed a patent convering various applications of metadata back in December, 2005. Specifically, two components in their abstract may affect our work here: first, syndication feeds for individual files or resources, and second, data repositories offering feeds about their content. Troubling. This is not the only one they recently applied for in this domain. I recall seeing a patent application for a system that uses pattern recognition on pictures, videos, and music to generate metadata, unfortunately I cannot find a reference. --[Si]dragon 22:27, 27 August 2006 (EDT)
  • It is better not to study existing patents when doing this sort of work. Let us end discussion here.
    1. It taints your mind with pre-conceived notions of how certain technologies should work.
    2. We're talking about ideas that are so far above the current state of the art that they probably "works around" any patents that may exist.
    3. Keep in mind that most software patents wouldn't stand up in court anyhow because they are too obvious. Honestly, you can't do any work in software today without unknowingly bumping into some silly patent. It's not worth worrying about. Fortunately, it's basically a "cold war." Everyone knows that if everyone started trying to enforce their software patents, the industry would be destroyed overnight.
    4. Our work is highly academic, not a commercial product. We should not let exterior legal nonsense affect our pure research. That is for those in business to hassle with.
    • Understood and agreed. Fortunately I only read the weblog which copied portions of the abstracts! --[Si]dragon 08:37, 28 August 2006 (EDT)

Generated metadata

• Is generated metadata such as derived from pattern recognition arbitrary or structured? --[Si]dragon 16:04, 28 August 2006 (EDT)
  • "arbitrary" and "structured" metadata are weak terms. We need to be more formal here. I just kinda made them up, for lack of better wording at the time. The general idea was that arbitrary metadata doesn't fit into a machine understandable schema or ontology. What is the meaning of a stand-alone tag called "vacation photos," for instance? That's arbitrary metadata. Off hand, I'm going to have to say that arbitrary metadata is nearly worthless.
  • Machine-generated metadata would definitely be structured because pattern recognition implies very strong context, unless you are talking about simplistic cases.

Need for clarification in terminology

We need to formalize our terminology before continuing this discussion. It makes little sense to discuss semantics without using standard semantics. Unfortunately, given that this is a cutting edge field, certain terminology seems to have loose definition. Some particular points: (references needed)

• Metadata is any data that describes data. However, this is relative to the specific usage. A poem by itself is data; as lyrics related to a song, it is metadata.
• Ontologies and schemas are metadata by some definitions, though they are rarely called this in practice.
• Ontology should generally refer to a formal ontology, as describable by an ontology representation language.
• We should prefer the term Schema for non-ontological but still hierarchical data models.

to be continued..