The Greater New York Area DB/IR Day will bring together database and information retrieval researchers and students from academic and research institutions across the tristate area and beyond for an exciting technical workshop and the opportunity to network.

The DB/IR Day will be hosted by Rutgers University and AT&T on Friday, April 28, 2006 from 9:45 am to 5:01 pm.

Provisional schedule:

Invited Talks:

Dennis Shasha (NYU)

Safe Data Management with Untrusted Servers

Imagine that you and your friends want to share information in a database because you want concurrency control, recovery, and query processing, but you don't trust the database administrator. You want to protect data from being observed (privacy). You want to make unauthorized modifications evident (a form of safety). You want to force the server to deliver a consistent picture to all honest users or be discovered (a form of liveness). Encryption and signatures make the first two possible. Liveness is another matter since the database administrator could "fork" the database into several copies, keeping some of your friends ignorant of your latest updates and you ignorant of theirs. In joint work with David Mazieres and some great students, we have worked out how to achieve these properties for file systems. This talk presents a design for database systems that integrates these goals with query processing, concurrency control, and recovery.

David Karger (MIT)

Why everyone should be their own database administrator, UI designer,
application developer, and web site builder, and how they can (PDF)

The desktop applications and web sites we use often seem ill-matched to our particular information management tasks. They fail to display, or even to record, some aspect of the information that we need. They clutter their presentations with distracting inessential aspects.
Information is fragmented over multiple applications and sites, making it hard to record, visualize, or navigate important connections. The operation we want to apply to data locked inside
some web site is only available at another site, or inside one of our applications.

End users can fix many of these problems themselves, if they are given the right levers for reshaping information management tools and repositories to suit their needs. The Haystack system offers three such levers: a simple, structured-but-sloppy, information model for
holding whatever information a particular user considers important; a user-interface framework that can flex to present that kind of information; and tools that let end users "edit" (rather than program) and share information views and workspaces that are appropriate for the data they want to record and the tasks they want to perform. Our approach offers a way to build personalized information management applications for users' own data, and to create useful aggregations and visualizations of information dispersed over the standard and Semantic Web.

William Cohen (CMU)

On Beyond Hypertext: Searching in Graphs Containing Documents, Words, and Actual Data (PDF)

Similarity measures for text have historically been an important tool for solving information retrieval problems. Here I will describe similarity metrics for graphs containing a heterogeneous mixture of textual and non-textual objects. These similarity metrics are based on a lazy graph walk, and they allow certain types of queries to be easily formulated by a naive user. In one instantiation of this framework, a user's personal information is represented as a graph containing messages, calendar information, social network information,
and a timeline. Here graph-based similarity search can be used to find identifiers for people mentioned in an email, or people likely to attend a meeting. In another instantiation, a graph is built that contains Medline abstracts, the output of a gene-name recognizer on these abstracts, a dictionary of gene synonyms, and previously curated abstracts. Here graph-based search can be used to find identifiers of genes mentioned in an abstract. In each of these cases, performance of the graph-based similarity search is competitive with or better than baseline approaches to the problem. In many cases, performance can be further improved by using appropriate learning techniques.

This is joint work with Einat Minkov and Andrew Ng.

Organization Committee:

• Amelie Marian (CS, Rutgers)
• Gheorghe Muresan (SCILS, Rutgers)
• Muthu Muthukrishnan (CS, Rutgers)
• Divesh Srivastava (AT&T)

dbir2006

sakai.rutgers.edu

Accommodation

If you intend to stay in New Brunswick overnight, before and/or after the workshop, a couple of hotels to consider are:

• Hyatt
  • This is probably more convenient: it is right next to the New Brunswick train station and the downtown / theatre area. To get to the Busch Campus (the Fiber Optic building is behind the Busch Student Center), one can use a Rutgers bus: walk one or two blocks into the College Avenue Campus, across/under the railway line, and take a bus A to the Busch Campus. Even better, take a cab for ~$10.
  • Unless your institution has a better deal, you should ask for the Rutgers rate.
• University Inn
  • This is a bit cheaper and in a more scenic location, but a bit more difficult to get to. You can get there from George St, next to the train station, by taking the Rutgers bus F to Douglass / Cook Campus. From there you can get to the workshop location by taking bus EE or F to College Avenue Campus, then the A bus to the Busch Campus. Even better, take a cab.
    

Posters

Previous DB/IR Days:

• Columbia University, April 15, 2005
• University of Pennsylvania, October 28, 2005