Tuesday, October 03, 2006

DBE Recommender Visualisations (V.0000001)


I have been working/thinking towards a way to represent service recommendations in a Service Oriented Architecture (SOA) . The visualisation tries to be exploratory, as a way to experience the "service landscape" available given a user profile/user history and current business processes. Currently there is a growing interest on recommenders for online systems, whatever their nature. Think of Amazon and their "Customers that bought this....".

Within the DBE context, the role of the recommender is to tell you which services could be a good match given your preferences and your runtime history. However and not surprisingly, it turns out that doing Business Process matching in an ad-hoc basis is hard stuff [1] at the technical service description level. The lack of semantics do not help either [2].

Reading some of the literature regarding SOAs, I have discovered that someone [3] came with a list of what the features of SOAs:

- Services are discoverable and dynamically bound

- Services are self-contained and modular

- Services stress interoperability

- Services are loosely coupled

- Services have a network-addressable interface

- Services have coarse-grained interfaces

- Services are location-transparent

- Services are composable

- Service-oriented architecture supports self-healing

Saying all this is easier said that done, specially having something that will work across organisations (where is public UDDI now? apparently someone pulled the plug!) rather than just one place/one organisation.

The idea behind the visualisation is to let the user play and quickly interact with service landscapes while making visible implicitly / explicitly the different SOA properties of the services . That’s why it is Version 0.00001 and there are still lots of work to be done.

Also, the user interface should enable the user to discover new services that might be relevant for a need that the user does not know he/she has yet, serendipity style. By enabling exploring/serendipity, it is possible that users discover new services that might be useful to improve a business process or to create a new business opportunities.

The principal idea behind the graphics is that if an SME is displayed at the centre, two rings could represent the service landscape, from the current services (one ring) to other recommended services. By chance, this representation allows you to see the services and the SME icon on the middle as a buttefly (left wing/right wing) with a bit of imagination.

The data was generated from a simulation that used Collaborative Filtering [4] techniques using the Taste toolkit[5] from a dataset that approximates SME behaviour, and the interaction design and visuals are done with Prefuse[6].

See the PDF for details:


[1] B. Mahleko, A.Wombacher, P.Fankhauser, “Process Annotated Service
Discovery Facilitated by an n-Gram Based Index”, in proceedings of
IEEE International Conference on e-Technology, e-Commerce and e-
Service (EEE), 2005, pages 2–8.

[2] R. Krummenacher, M. Hepp, A. Polleres, C. Bussler and D. Fensel,
“WWW or What Is Wrong with Web Services”, in proceedings of the
2005 IEEE European Conf on Web Services, 2005.

[3] J.Borenstein and J. Fox, “Enterprise Semantics: Aligning Service-
Oriented Architecture with the Business”, WebServices Journal,vol. 5,
no.4, 2005, pages 30-34.

[4] J. L. Herlocker , J. A. Konstan, L. G. Terveen University of Minnesota, Minneapolis, MN, John T. Riedl University of Minnesota, Minneapolis, Evaluating collaborative filtering recommender systems ACM Transactions on Information Systems (TOIS) a Volume 22 , Issue 1 (January 2004) Pages: 5 - 53

[5] Taste. Collaborative Filtering for Java. http://taste.sourceforge.net/

[6] Prefuse. Interactive Information Visualisation Toolkit. http://prefuse.org/



Post a Comment

<< Home