Most information access tools are designed for ad-hoc, single-use interactions with anonymous individuals. However, people now use computers routinely, for a wide variety of tasks, over multi-year periods. People also assemble ad-hoc personal digital libraries that they carry with them from one computer to another, across a lifetime of use. There is growing recognition that the next generation of information access tools must make greater use of context, in the form of detailed user and task models, to provide greater personalization and improved accuracy. This session will focus primarily on use of unstructured and semi-structured data.

In the past few years we have seen significant progress on many aspects of data integration. The community has developed flexible architectures for data integration, powerful methods for mediating between disparate data sources, tools for rapid wrapping of data sources, methods for optimizing queries across multiple data sources, and more. In parallel, some of the data integration innovations have been converted to commercial systems.

This breakout session will discuss the agenda for future research on data integration. Examples of some of the issues we'll be addressing are:

1. What have been the main successes of data integration and what have we learned from them?
2. What are the key challenges going forward?
3. What are possible approaches for dealing with semantic heterogeneity?
4. What are the relationships between the Semantic Web and Data Integration? Can the two technologies support each other?
5. What are the cross-disciplinary opportunities to be explored in this realm between the DB, AI and IR communities?
6. Can we come up with agreed upon benchmarks for research on data integration?
                                               

The field of medicine has become increasingly information and data intensive. Medical researchers, clinicians, and patients all face challenges in managing vast and expansive sources of data and information to accomplish their desired tasks. In this breakout group, we will examine these challenges with respect to the overall workshop theme: synergies and synthesis between information retrieval and databases. Example questions to discuss are: 

• What challenges facing researchers, clinicians, or patients could benefit from synergies between information retrieval and databases?

• What are the best avenues for creating these synergies among the medical informatics, information retrieval, and database communities?

• Focusing on our own research agendas, what benefits do we see for promoting such synergies?

Many important information systems problems primarily involve the manipulation of structural meta data, that is, models (e.g. schemas, interfaces, web-site maps, etc.) and mappings between modes. Examples include schema evolution, XML message translation, application integration, data warehouse loading, database wrapper generation, and design tool implementation. Despite the similarity of solutions to these problems, today they are solved in an application-specific way and usually require much object-at-a-time programming.

The goal of model management is to develop a generic infrastructure that offers a major productivity improvement to builders of such model-driven applications. The main abstractions are models and mappings between models. It treats these abstractions as bulk objects and offers such operators as Match, Merge, Diff, Compose, Extract, and ModelGen. An overview of model management appears in: http://www-db.cs.wisc.edu/cidr/program/p19.pdf

Although work meta data problems is as old as the database field itself, work on generic solutions have only recently become popular. Thus, there are many research opportunities. A discussion of future research could include the following issues:

1. What kinds of application problems should be addressed? How does this address problems in currently hot areas, such as bioinformatics and the semantic web?
2. What are good choices for representing the semantics of mappings? What past work on mapping languages could be reused in this context?
3. User interfaces are extremely important when manipulating large models and mappings. How can we improve them?
4. Traditionally, meta data management is treated as a database problem. However, recently machine learning has been successfully applied to the schema matching problem. How can other non-database techniques should be applied to model management? E.g., data mining, theorem proving, text retrieval, natural language understanding, combinatorial optimization.

The phrase Personal Information Management or PIM refers not so much to an established field of inquiry as it does to a gnawing, nagging need. It's a need we all share but that we each must meet in our own individual, idiosyncratic ways. Now is a good time to take stock of PIM as a need and as a field of inquiry with special focus on the ways in which computing technology can help. The panel will address the following questions:

• What is PIM as a field of inquiry? What should it encompass?
• In what ways has computing technology helped? In what ways has it hurt?
• Panelists are encouraged to bring and read their own favorite samples from the sometimes breathless predictions of twenty years back. Some dreams have come true. Others have not. Some may now seem wildly off base. Why? What can we learn from this going forward?
• What have we learned in the past twenty years through the study of PIM? What do we know through empirical inquiry? What have we learned through tool prototyping?
• What don't we know? What should we be finding out?

Lack of trust, security, and reliability impedes information sharing. Potential for theft, fraud, harassment and destruction of critical private data exists despite an increasing focus on security. The session will address the following challenges:

• Computational models to formalize trust and fraud and for analyzing vulnerabilities and threats in cyberspace
• Privacy preserving access to databases over trusted communication
• Authorization based on evidence and trust
• Formalize evidence
• Recommendation and reputation systems
• Experiments, prototypes and applications
• Relationship with other scientific disciplines
• Participants are encouraged to propose ideas used in other domains of computer science or other fields
• Activities in industry
• We need to determine the current status of research and propose plans for future

Please send any questions or problems on this www page to:  nsf2003@cs.washington.edu

Mon May 19 16:33:53 PDT 2003