Research areas

The laboratory for web science (LWS) focuses on three research areas: Complex Networks, Recommender Systems and Semantic Web. Results from all areas find applications in industry projects. 

Complex Networks

Most technical, sociological and biological networks reveal topological patterns, that are neither purely random nor regular. Their degree distribution have fat tails and non trivial features like high clustering coefficients are observed. These properties have a big impact on dynamical processes taking place on such networks.

The laboratory for web science analyzes complex networks by means of modeling and simulations of dynamical processes. Results are compared to measurements from real world networks like the Internet, WWW, collaborative graphs and others. Furthermore we focus our research on spectral graph algorithms to detect non trivial clusters in complex networks.

A real world example of a complex network (collaboration graph) is shown below:

Recommender Systems

As an information society we are faced with the problem of information overflow. How to find customized information in a huge network like the World Wide Web?

Search Engines are hard workers to help but they have some weaknesses. They are not (yet) able to deliver information according to particular preferences.

Recommender systems are one possible answer to this problem. They are able to 'learn' user preferences and to filter information according this knowledge. In addition they are smart enough to recommend information/objects to users. Probably the most popular recommender is driven by amazon. Recommender systems are classified in three main groups:

Content based: Commonalities between already consumed objects/information in the past are analyzed and items with similar features are recommended to users.

Collaborative filtering: The system takes into account consumer behavior of all participants and calculates a user-user or object-object network. The system then recommends objects/information based on the generated network.

Network based: These algorithms omit the projection process - bipartite graph to unipartite graph -  and analyze the user-object network directly. To explore the structure, diffusion like processes are simulated on the network. See B-Rank: A top N Recommendation Algorithm as an example.

Data visualization of a real recommender system (movielens.org):

Movie recommender system projected as a user-user network.

There are many applications for recommender systems. From online platforms to expert systems in medical diagnostics.

The laboratory for web science develops new algorithms and verifies their strength in industry projects. In addition we develop theoretical models of user behavior to better understand what algorithm fits best a given problem set (data generated by users).

Semantic Web

The semantic web - envisioned by Tim-Berners Lee - is the attempt to formalize concepts and relationships within a given knowledge domain. This allows a program to understand the semantic of data. Concepts are realized as ontologies using a formalism like OWL. 

The laboratory for web science actively researches on ontology development and the convergence of the Web2.0 Ansatz with ontology based concepts. This is called Web3.0 sometimes. Furthermore we develop protocols for autonomous agents based on communication ontologies like FIPA.

All research areas have a substantial part in education and we emphasize on the integration of research results into education.