Projects

SWARMD is a shortcut for Specialized Web Agents to Relieve Man-Mashine Dialog. SWARMD is an extensive system for personalized information-retrieval, information-analysis and information-representation.

Motivation: The search for personalized and specialized information becomes more and more complex for organsiations and people as well. This asks for more resources - time and money.
Search engines like goolge, clusty, a9 and others are based 'on wisdom of the crowds' and blind for individual preferences in general. SWARMD helps to fill this gap. The system searches for information based on individual profiles, analyses the data and presents results in a context network.

SWARMD is based on programmable, autonomous agents, which independently search for desired information in a knowledge network like the WWW. Agents follow a evolutionary paradigm: an agent resembles a species with particular properties. Mutation is mimiced by punctual pertubation of a random selected property. Selection is driven by users feedback: if an agent is effective in finding and analyzing information he gets a higher fitness. Agents less successful are punished by a decrease in their fitness and may die.

SWARMD is a new paradigm: turning search tasks to recommendation tasks.

SWARMD is a project together with the University of Applied Sciences in Chur (HTW).

Contact:Dr. Marcel Blattner

The laboratory for web science developed a very promising algorithm for recommender systems. The algorithm has a high recall, precision rate and keeps a high inter-list diversity as well.
The performance improvement is about 60% compared to other benchmark algorithms. The algorithm is parameter-free which makes it very attractive for industry.

B-Rank is a core component in a proposed CTI project. The project is evaluated by the CTI these days.

Publication:

http://www.fernfachhochschule.ch/ffhs/afe/lws/forschung/publikationen

Contact:Dr. Marcel Blattner