Tuesday, 5/22/2007
4:00 PM - 4:30 PM
Level: Technical - Introductory

Creating a smarter version of the web, a web that encodes meaning, is still far from reality on a broader scale and many open questions remain despite all suggested standards and technologies. Many approaches for building the semantic web are today focused entirely on ontologies and encoding of relatively simple metadata, considering inference as an add-on feature when rich data is in place. The problem is that inference is directly connected to and affected by the underlying representation, meaning that many design decisions are done blindfolded, resulting in difficult limitations when moving towards more reasoning functionality over time. This talk introduces a practical approach to building semantic web applications with main focus on inference, an approach quickly resulting in smart applications that can be built and maintained by a diverse audience with domain knowledge. This approach relies on machine learning, case-based knowledge representation, multiple approaches for inference and data mining instead of “conventional” semantic web standards and tools.

• Comparisons between the different approaches using practical examples
• Demonstration of a wide range of semantic applications, from vertical search to biotech
• Intelligent resource linking and trust creation networks

Lars has over 20 years of industrial experience in the video games industry and applied Artificial Intelligence and Computational Intelligence in a wide range of domains, from evolutionary social gravitation models to advanced self-learning systems for handling a large volume of support requests for telecom corporations.