Challenges set by the Urban Computing use case to other LarKC WPs

TBC!!!

challenges for LarKC:
- elimination of duplicates in event description:
  - events info are provided in a "wiki" approach so it's possible to have duplicates
  - rule-based approaches to eliminate duplicates? (e.g. same place same time and similar description means same event)
  - machine learning trained on manually-detected duplicates?
- disambiguation of geographical names:
  - e.g. there are several places called Milano around the world (see here)
  - paper by Raphael Volz & Co. about using ontologies to disambiguate
- machine learning challenge for traffic prediction?
- Hybrid components and reasoner(?)
  - ability to process any reasoning requirement such as forward and backward reasoning, DL and Rule based reasoning, static and dynamic reasoning.
  - Related to Frank’s email
    Example5 (hybrid components): We should definitely investigate closer integration of (say) abstraction + reasoning or (say) selection + reasoning. This has been strongly initiated at the Berlin meeting by Dieter. But this can be easily done exploiting the infrastructure of the current LarKC architecture, since the boundaries between the different LarKC plugin-types are sufficiently fluid. (this in line with comments by Michael and Georgina)
- parallel computing/cloud computing
  - ability to process huge number of data within quasi-real time.
  - Example1 (Big Hammer): A highly parallelised cluster-based inference engine (where all the data is stored only once, inside the cluster) can be combined with components that select only a small subset of the resulting triples for further shipping between components. (work by VUA and Stuttgart). This is scalability with a Big Hammer, and currently LarKC has the world's best results in this area (with even better ones to come, watch the ESWC2010 submission deadline).