---------------------------------------- ----------------------------------------

Workshop on Link Analysis: Dynamics and Static of Large Networks (LinkKDD2006)

August 20, 2006

---------------------------------------- ----------------------------------------

To be held at KDD-2006, The Twelfth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Philadelphia, PA, USA, 20-23 August 2006

Call for papers (expired) on the following topics was open until June 12, 2006. ----------------------------------------

In the past years there has been great interest in developing information technology for Link Discovery (LD). LD is relevant to a wide range of research topics that has been developed in past decades. LD has its roots in various fields such as discreet mathematics, graph theory, social science, pattern analysis, link analysis and spatial databases. The common goal of this research is the development of techniques for mining large collections of data to extract valuable knowledge that may be present as hidden patterns or links among seemingly unrelated items. Successful LD applications will discover the hidden structure of organizations, relate groups, identify fraudulent behavior, model group activity and provide early detection of emerging threats. The broader context of the workshop can be related in some respect to the areas of Data Mining, Machine Learning, Information Retrieval, Natural Language Processing, Social Networks Analysis, and the general Graph Theory.

LD requires a radically different approach to knowledge discovery, both in techniques and in approaches to evaluating LD algorithm results. The departure from standard approaches is made clear in the following five characteristics of LD problems and their representation: 1) Data is heterogeneous, arriving from multiple sources. The data and patterns sought include representations of people, organizations, objects, actions and events. Each of these entities has its own set of attributes, and there are many types of relations that might exist between them. 2) Unlike conventional data mining, in which nodes are variables and links are statistical relations among variables, nodes represent entities and links are relations amongst entities. 3) LD assesses the likelihood that an instance of a specific graph-theoretic structure in the data matches a pattern of interest. The structure may include temporal, spatial, organizational, and/or transactional patterns. 4) All LD problems involve estimating a population based on a sample of data. Typically, a relatively low number of observations for each entity can be recorded, and the overall sample is typically small relative to the size of the population. 5) The data becomes available over time, so the timing of when to make a decision based on LD analysis is a central issue.

Although LD and group detection attracted researchers' attention, this is still an open research issue. We believe that much progress could be achieved from a concerted effort and a greater amount of interactions between researchers in all these different research areas. The purpose of this workshop is to provide a forum to foster such interactions, discuss the new achievements and identify future research directions.

We believe that much progress could be achieved from a concerted effort and a greater amount of interactions between researchers in all these different research areas. The purpose of this workshop is to provide a forum to foster such interactions, discuss the new achievements and identify future research directions. ----------------------------------------

Topics of interest

Particular topics of interest for the workshop include but are not limited to:

----------------------------------------

Workshop Program

The workshop consists of invited talk, presentation of refereed papers, and discussions. We hope that the program will stimulate future collaboration among researchers ----------------------------------------

Attendance

Attendance is not limited to the paper authors. We strongly encourage interested researchers from related areas to attend the workshop. The workshop should be of interest to researchers and practitioners conducting research or building applications that involve various data analysis and rich data and knowledge representations, in particular, those from:

We expect that the workshop topic will attract attention of regular KDD attendees who are interested in Data Mining, Machine Learning, and Natural Language Processing but also potentially encourage the attendance of KDD by participants interested in Social Network Analysis who would otherwise not choose to come to the conference.

----------------------------------------

Organization

Program Chairs

Marko Grobelnik
J.Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia

Jafar Adibi
University of Southern California, 4676 Admiralty Way Suit 1001, Marina del Rey, CA 90292

Natasa Milic-Frayling
Microsoft Research Ltd, 7 J J Thomson Avenue, Cambridge, CB3 0FB, United Kingdom

Dunja Mladenic
J.Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia

Patrick Pantel
University of Southern California, 4676 Admiralty Way Suit 1001, Marina del Rey, CA 90292

Program Committee (to be confirmed)

----------------------------------------

Past events

We feel that the continuity of meeting and exchanging ideas is essential for effective promotion and development of this research area.

----------------------------------------