Linked Data - Part 1 to 6

Summary

In part 1 we talked about Linked Data and noted that it's a good idea, but that it wasn't suited for something like a knowledge-database. In part 2 we talked about how normal companies have no real incentive to store their 'knowledge' in Linked Data's RDF data format, because the biggest advantage of linked data is for other people and not for the company itself. In part 3 we finally started to think about what kind of database would be suitable for storing knowledge (even for normal companies) and we thought of some overly ambitious goals. We started with one of the goals, which is: "Allowing author-defined structure to our unstructured database". So if, for example, the creator of the database wants a tabular-like structure, we can easily enforce it. In part 4 we figured out that the typed relations as used in Linked Data's RDF triplet-store isn't all that great and that we are probably better off using a normal directed graph. In part 5 we've shown that author-defined structure could even enforce a triplet-like structure as in RDF while still using a simple directed graph. Our biggest problem was that we needed to create patterns to find matches in the database which is basically a type of subgraph isomorphism problem (an NP-complete problem). We tried to brainstorm on using a caching-method for pattern matching so that we don't have to run a complex algorithm on every graph-modification. In part 6 we finally looked at the pattern-match problem itself without looking at the "author-defined consistency rules" and ignored the caching-idea for a while.

So we went from a very broad idea with very ambitious goals to a small, tangible, sub-problem. This is a good thing because we can't just keep dreaming of things that should have been, without giving a prudent alternative. So our focus is currently:

  1. Create an algorithm for pattern-matching (that hopefully can also be used for querying).
  2. Support caching for pattern-matching.
  3. Use pattern matching (with caching) to construct author-defined consistency rules to enforce additional structure.
  4. Find a new tangible sub-problem that is fun to work on.
     

line